MEASURING ORGAN DOSES AND ASSESSING CLINICAL IMAGE QUALITY FOR THE PURPOSE OF COMPUTED TOMOGRAPHY PROTOCOL OPTIMIZATION

Transcription

1 MEASURING ORGAN DOSES AND ASSESSING CLINICAL IMAGE QUALITY FOR THE PURPOSE OF COMPUTED TOMOGRAPHY PROTOCOL OPTIMIZATION By IZABELLA L. LIPNHARSKI A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA

2 2017 Izabella L. Lipnharski 2

3 To my loving family 3

4 ACKNOWLEDGMENTS The completion of the work in this dissertation would not have been possible without the support of several amazing individuals. I owe my deepest gratitude to my advisor, Dr. Arreola, for his precious guidance, insight, and contagious enthusiasm throughout my graduate career. Dr. Arreola s unending patience and persistent motivation led me through several challenging and rewarding experiences that would not have been possible without him. I thank him for the incredible opportunity to complete such meaningful research and for allowing me to pursue a career I am head over heels passionate about. I am forever grateful for his kindness towards his graduate students, always reminding us that life comes first. Secondly, I am grateful for having had such inspirational figures in my academic and professional life, especially Dr. Rill, Dr. Brateman, and Dr. Leon. Their leadership and strict adherence to excellence has had a great impact on the way I conduct my work. I look up to these brilliant women with great respect and will always have the fulfillment of knowing I had the pleasure of learning and growing with their help as teachers and mentors. I d like to thank Dr. Hintenlang for his support during my graduate career and for his valuable input as my supervisory committee member. His thoughtful feedback improved my research approaches and helped me develop stronger problem solving skills. I would also like to thank Dr. Entezari for his insight and encouragement as my external committee member. I am grateful for his time and honest interest in this research. I am especially thankful for the radiologists whose sincere interest, support, hard work, and direction drove the bulk of this research. Dr. Mohammed, Dr. Verma, and Dr. 4

5 Rajderkar provided endless hours of advice, read several postmortem CT images, and helped recruit additional radiologists for our work. Several other radiologists in the department also contributed their valuable time to evaluate images for improving patient care. This work would not have reached completion without the help and interest of three wonderful research assistants. Catherine Carranza, Nathan Quails, and Nathalie Correa are some of the most hard-working colleagues I have worked with. They made me look forward to coming into the office, always knowing that they would have new updates about science, life, or a new joke to tell, and were always willing to help when the work was overflowing. I am extremely proud to see each of them flourish and grow with knowledge of imaging systems and take on complex tasks independently. I thank Anna Mench and Becky Lamoureux for teaching me the foundation of organ dosimetry and for welcoming me with open arms to the team. I am grateful for watching them move on to their careers and share their experiences with me. Several residents took their time to impart a fraction of their knowledge onto me, and I am grateful for Michael Wayson, Zemei Lui, Matt Hoerner, Donglai Huo, and BC Schwarz. I am grateful for the support of my excellent professors at the University of Miami. I thank Dr. Manns for opening my eyes to the field of medical imaging, Dr. Zhao for entertaining several conversations regarding a career in medical physics, and Dr. Bohorquez for his passionate teaching and guidance. They have been instrumental to my success and I am happy to keep in touch several years later. I owe a deep thank you to my remarkable friends who filled my life with happiness and created memories worth treasuring. They have supported me in the 5

6 midst of anxiety and stress, added value to my thinking, and reminded me that I am bigger than my thesis. Their moral support has made all the difference. There are no words to describe my love and gratitude to my amazing family. My parents have made a huge impact on my academic achievements and I am truly thankful for their immeasurable sacrifice, guidance, and unconditional love. I am beyond thankful to my father for instilling the curiosity to pursue challenging work and the drive to be successful at it. My mother s exemplary work ethic, positive attitude, and selfless love has always inspired me. I learned from her that difficult times serve to test your spirit and that hard work always pays off. I thank my younger brother for always being there to encourage me in my personal life and for shaping the person I am today. My parent in-laws have also provided great conversations and encouragement from the first day I met them. I am grateful for their love and genuine interest in my life. Finally, and above all, I cannot begin to express my unfailing gratitude and love to my husband, Daniel. I thank him for making me smile and giving me himself to look forward to, knowing that at the end of the day, his persistent smile, sunny optimism, gentle patience, and unwavering love would always keep me going in the right direction. I thank him for always encouraging me when the tasks seemed arduous and insurmountable. I am grateful to Daniel not just because he has given up so much to make my career a priority in our lives, but because he has seen me, and loved me, through the ups and downs of the entire academic process. He has shared this entire amazing journey with me. 6

7 TABLE OF CONTENTS page ACKNOWLEDGMENTS... 4 LIST OF TABLES LIST OF FIGURES LIST OF ABBREVIATIONS ABSTRACT CHAPTER 1 INTRODUCTION Growth of Computed Tomography Increase of Radiation Exposure Radiation Risks Need for Accurate Organ Doses Protocol Optimization in Computed Tomography Specific Aims PRINCIPLES OF CT TECHNOLOGY A Brief History of CT Development of CT Technology Helical Scanning Helical pitch Multiplanar imaging Multiple-Detector CT Dose Reducing Technologies Tube Current Modulation Iterative Reconstruction Filtered backprojection Statistical iterative reconstruction Adaptive iterative reconstruction Organ Shielding DOSIMETRY IN COMPUTED TOMOGRAPHY Multiple Scan Average Dose (MSAD) CT Dose Index (CTDI) CTDIFDA CTDI CTDIw CTDIvol

8 3.3 Dose Length Product (DLP) Effective Dose Limitations of Dose Metrics CT ORGAN DOSE MEASUREMENTS IN CADAVERS Cadavers Utilized in this Work Organ Dose Measurements Optically Stimulated Luminescent Dosimeters Organ Dose Measurement Methodology CT Scanner Utilized in this Work Automatic Exposure Control Utilized in this Work Iterative Reconstruction Utilized in this Work CT Scan Protocols ASSESSING CLINICAL IMAGE QUALITY FOR PROTOCOL OPTIMIZATION Image Quality in Computed Tomography Measuring Quantitative Image Quality CT number accuracy and uniformity High-contrast resolution Low-contrast resolution Evaluating Qualitative Image Quality Receiver operator characteristic analysis Visual grading analysis Image Quality Evaluation Method Used in this Work Task-Based Image Quality Analysis Ideal Grading Scale for Protocol Optimization Statistical Analysis Identifying Clinical CT Exam Indications for Protocol Optimization Assessing the Usefulness of Postmortem Subjects in CT Image Quality Methods Results Creating Indication-Specific Image Quality Questionnaires Methods Results Conducting an Observer Study Methods Results Discussion PROTOCOL OPTIMIZATION IN CT OF PATIENTS WITH METAL IMPLANTS Presence of Metal in CT Imaging: A Retrospective Analysis at UF Health Methods Results Effect of Metal Implants on Patient Dose

9 6.2.1 Methods Results Effect of Metal Implants on Image Quality Methods Scanning with a CAP protocol Scanning with a pelvis protocol Results Scanning with a CAP protocol Scanning with a pelvis protocol Effect of a Metal Artifact Reduction Algorithm on Image Quality Methods Scan techniques for subject Scan techniques for subject Image quality evaluation Results Image quality scores for subject Image quality scores for subject SEMAR s Performance with Varied kvp and ma Methods Scan techniques for subject Scan techniques for subject Image quality evaluation Results Image quality scores for subject Image quality scores for subject Discussion PROTOCOL OPTIMIZATION IN LUNG CANCER SCREENING CT Lung Cancer Screening in CT Measuring Organ Doses from Lung Cancer Screening CT Methods Results Optimizing Image Reconstruction in Lung Cancer Screening CT Methods Results Assessing Image Quality for Reduced-Dose Lung Cancer Screening CT Methods Results Measuring Organ Doses from Reduced-Dose Lung Cancer Screening CT Methods Results Analysis of Patients Examined with Lung Cancer Screening CT at UF Health Methods Results Discussion

10 8 PROTOCOL OPTIMIZATION IN PEDIATRIC BODY CT Pediatric CT Imaging Validation of a Postmortem Subject as a Surrogate for a Pediatric Child Methods Results Measuring Organ Doses from Pediatric Body CT Methods Results Measuring Organ Doses from Utilizing Current Dose Reduction Methods Appropriateness of CT Examination Child-Sized Techniques Iterative Reconstruction Organ shielding Methods Results Assessing Image Quality for Reduced-Dose Pediatric CT Exams Methods Results Measuring Organ Doses from Reduced-Dose Pediatric CT Exams Methods Results Discussion PROTOCOL OPTIMIZATION IN PEDIATRIC HEAD CT Measuring Organ Doses from Pediatric Head CT Methods Results Assessing Image Quality for Reduced-Dose Pediatric Head CT Indications that Warrant Dose Reduction Hydrocephalus Ventriculoperitoneal shunt Trauma follow-up Craniosynostosis Methods Results Measuring Organ Doses from Reduced-Dose Pediatric Head CT Methods Results Analysis of Pediatric Patients Examined with Head CT at UF Health Methods Results Discussion CONCLUSION

11 10.1 Summary of Findings Metal Implants in CT Lung Cancer Screening Pediatric Body CT Pediatric Head CT Organ doses General Recommendations for Protocol Optimization Future Work Final Words LIST OF REFERENCES BIOGRAPHICAL SKETCH

12 LIST OF TABLES Table page 4-1 Height, weight, BMI, and effective diameters of cadavers utilized in this work Number of OSLDs utilized per organ for each Postmortem Subject Scan techniques for standard CAP protocol utilized on Subjects 1, 2, and Organ doses measured with CAP CT scans of Subject 1, 2, and Scan techniques for standard adult CAP protocol utilized on Subject Radiologists utilized to assess image quality of a CAP CT scan of Subject Scan techniques for standard adult PWH protocol utilized on Subject Radiologists utilized to assess image quality of a PWH CT scan of Subject Scan techniques for standard adult PWH protocol utilized on Subject Radiologists utilized to assess image quality of a PWH CT scan of Subject Image quality scores of Subject 3 scanned with a CAP CT protocol Image quality scores of Subject 3 scanned with a PWH CT protocol in regions of the pelvis with and without metal implants Image quality scores of Subject 5 scanned with a PWH CT protocol in regions of the pelvis with and without metal implants Image quality scores of Subject 3 scanned with a PWH CT protocol with and without SEMAR Image quality scores of Subject 5 scanned with a PWH CT protocol without and with SEMAR Scan techniques for modified PWH protocols acquired with metal artifact reduction on Subject Scan techniques for modified PWH protocols acquired with metal artifact reduction on Subject Image quality scores of Subject 3 scanned with modified PWH CT protocols with and without SEMAR Image quality scores of Subject 5 scanned with modified PWH CT protocols without and with SEMAR

13 7-1 Organ doses measured in Subject 5 scanned with a LDCT LCS protocol Organ doses measured in Subject 7 scanned with a LDCT LCS protocol Scan techniques utilized on Subject 3 scanned with various LDCT Lung Cancer Screening protocols for image quality assessment Scan techniques utilized on Subject 4 scanned with various LDCT Lung Cancer Screening protocols for image quality assessment Scan techniques utilized on Subject 5 scanned with various LDCT Lung Cancer Screening protocols for image quality assessment Scan techniques utilized on Subject 7 scanned with various LDCT Lung Cancer Screening protocols for image quality assessment Scan techniques utilized on Subject 8 scanned with various LDCT Lung Cancer Screening protocols for image quality assessment Scan techniques utilized on Subject 9 scanned with various LDCT Lung Cancer Screening protocols for image quality assessment Scan techniques utilized on Subject 10 scanned with various LDCT Lung Cancer Screening protocols for image quality assessment Scan techniques utilized on Subject 11 scanned with various LDCT Lung Cancer Screening protocols for image quality assessment Radiologists utilized to assess image quality of lung cancer screening scans Image quality scores for diagnosing lung cancer screening in Subject Image quality scores for diagnosing lung cancer screening in Subject Image quality scores for diagnosing lung cancer screening in Subject Image quality scores for diagnosing lung cancer screening in Subject Image quality scores for diagnosing lung cancer screening in Subject Image quality scores for diagnosing lung cancer screening in Subject Image quality scores for diagnosing lung cancer screening in Subject Image quality scores for diagnosing lung cancer screening in Subject Scan techniques utilized on Subject 5 scanned with various Low-Dose Lung Cancer Screening CT protocols for organ dose measurement

14 7-21 Scan techniques utilized on Subject 7 scanned with various Low-Dose Lung Cancer Screening CT protocols for organ dose measurement Organ doses measured in Subject 5 scanned with standard and reduceddose LDCT LCS protocols that produced accep image quality Organ doses measured in Subject 7 scanned with standard and reduceddose LDCT LCS protocols that produced accep image quality Percent organ dose reductions achieved in Subject 15 with reduced-dose LDCT LCS protocols that produced accep image quality Percent organ dose reductions achieved in Subject 17 with reduced-dose LDCT LCS protocols that produced accep image quality Scan techniques utilized on Subject 5 scanned with a standard pediatric chest CT protocol and a standard pediatric abdomen CT protocol Organ doses measured in Subject 5 scanned with a standard pediatric chest CT protocol and a standard pediatric abdomen CT protocol Scan techniques utilized on Subject 5 scanned with abdomen CT protocols using adult settings versus pediatric settings Scan techniques utilized on Subject 5 scanned with pediatric abdomen CT protocols acquired with and without iterative reconstruction algorithms Scan techniques utilized on Subject 5 scanned with pediatric chest CT protocols with and without breast bismuth shields Average organ doses and reductions from utilizing adult versus pediatric scan settings in an abdomen CT scan Average organ doses and reductions from utilizing iterative reconstruction versus filtered back projection in a pediatric abdomen scan Average breast dose and reduction from utilizing bismuth breast shields in a pediatric chest scan Scan techniques utilized on Subject 5 scanned with standard and reduceddose pediatric chest CT protocols Scan techniques utilized on Subject 5 scanned with standard and reduceddose pediatric abdomen CT protocols Radiologists utilized to assess image quality of pediatric CT scans Image quality scores for diagnosing pediatric chest CT protocols

15 8-13 Image quality scores for diagnosing pediatric abdomen CT protocols Organ doses measured in Subject 5 scanned with standard and reduceddose pediatric chest CT protocols that produced accep image quality Organ doses measured in Subject 5 scanned with standard and reduceddose pediatric abdomen CT protocols that produced accep image quality Maximum achievable organ dose reductions (%) from using pediatric chest and abdomen protocols that produced accep image quality Average pediatric chest CT organ doses measured in this work compared to organ doses calculated elsewhere Organ doses (mgy) measured for a standard Head CT protocol and Trauma Head and C-Spine CT protocol in Subject Radiologists utilized to assess image quality of pediatric CT scans Image quality scores for diagnosing hydrocephalus with 1 standard and 10 low-dose head CT protocols in Subject Image quality scores for diagnosing craniosynostosis with 1 standard and 10 low-dose head CT protocols in Subject Image quality scores for diagnosing trauma follow-up with 1 standard and 10 low-dose head CT protocols in Subject Image quality scores for diagnosing VP-shunt with 1 standard and 9 lowdose head CT protocols in Subject Organ doses measured for standard and low-dose Head CT protocols in Subject Organ doses measured for standard and low-dose Stealth Head CT protocols in Subject Clinical exam parameters for pediatric patients scanned with CT for craniosynostosis, hydrocephalus, and VP Shunt at UF Health Average head CT organ doses measured in this work compared to organ doses calculated elsewhere

16 LIST OF FIGURES Figure page 2-1 First clinical prototype EMI brain scanner First clinical image obtained from EMI CT scanner Illustration of principle of helical CT Illustration of principle of single slice CT and multi-detector CT Graph of tube current superimposed on a CT projection radiograph illustrating the concept of longitudinal dose modulation Graph of tube current superimposed on a CT projection radiograph illustrating the concept of angular-longitudinal dose modulation Illustration of principle of MSAD Equipment typically used to measure CTDI Photographs of nanodot, adapter and microstar reader Schematic of general access tube placement within the organs of interest Schematic of AIDR 3D process Prevalence of patients containing metal implants in 300 adults examined with CT at UF Health Types and number of metal implants encountered in 82 patients at UF Health Graphs of tube current modulating in Subjects 1, 2, and Average organ doses measured with CAP CT scans of Subjects 1, 2, and Image quality scores of Subject 3 scanned with a CAP CT protocol Image quality scores of Subject 3 scanned with a PWH CT scan Image quality scores of Subject 5 scanned with a PWH CT scan Pelvic images of Subject 3 reconstructed with and without SEMAR Pelvic images of Subject 5 reconstructed with and without SEMAR Image quality scores for pelvis scans of Subject 3 by image quality feature, radiologist ID, and metal artifact reduction

17 6-11 Image quality scores of Subject 5 scanned with a PWH CT scan by image quality feature and metal artifact reduction Image quality scores for soft tissue visualization of pelvis scans with and without SEMAR in Subject Image quality scores for bone detail of pelvis scans with and without SEMAR in Subject Image quality scores for PWH scans of Subject 5 by image quality feature, scan protocol, and metal artifact reduction Average organ doses measured in Subject 5 and Subject 7 scanned with LDCT LCS protocols Axial chest images reconstructed with the Standard Body Axial algorithm Axial lung images reconstructed with the Lung Sharp algorithm Axial lung images reconstructed with Lung Sharp, Lung Low Dose and Lung HRCT reconstruction algorithms Lung cancer screening CT images of Subject 5 acquired at 120 kvp with the protocol using the highest dose and lowest dose Lung cancer screening CT images of Subject 8 with the protocol using the highest dose and lowest dose Lung cancer screening CT images of Subject 9 with the protocol using the highest dose and lowest dose Image quality scores for diagnosing lung cancer screening with 8 CT protocols Average organ doses measured in Subject 5 scanned with LDCT lung cancer screening protocols that produced acceptable image quality Average organ doses measured in Subject 7 scanned with LDCT lung cancer screening protocols that produced acceptable image quality Plot of tube current modulating throughout the scan length of Subject Plot of tube current modulating throughout the scan length of Subject Histogram of patient weight for 109 patients examined with LDCT lung cancer screening examinations at UF Health Correlation of patient weight with BMI and effective diameter in the chest

18 7-15 CTDIvol and SSDE for 86 LDCT lung cancer screening patient examinations conducted on Siemens Sensation 16 CT Scanners CTDIvol and SSDE for 23 LDCT lung cancer screening patient examinations conducted on four Toshiba CT Scanners Average organ doses measured in Subject 5 scanned with a standard pediatric chest CT protocol and a standard pediatric abdomen CT protocol CTDIvol values achieved by increasing the target noise level values in pediatric chest and abdomen protocols Image quality scores for pediatric chest CT protocols Image quality scores for pediatric abdomen CT protocols Tube current modulation and average tube current Images of a PVC tube placed in third ventricle Images of a PVC tube and OSL dosimeters placed in the third ventricle Scan ranges for Head CT protocols at UF Health Screenshots of scan ranges for the Head protocol acquired on Subject Average organ doses measured for typical head CT protocols in Subject CTDIvol values from modifying scan parameters in Head protocols CTDIvol values from modifying scan parameters in Stealth protocols Graph of tube current of Head protocol with increasing target noise levels Graph of tube current of Stealth protocol with increasing target noise levels Graph of tube current of Head protocol with iterative reconstruction with increasing target noise levels Graph of tube current of Stealth CT protocol with iterative reconstruction with increasing target noise levels Graph of tube current of Head protocol with fixed tube current, tube current modulation, and iterative reconstruction with increasing target noise levels Graph of tube current of Stealth protocol demonstrating the effects of miscentered patients

19 9-14 Average organ doses measured from standard and acceptable reduced-dose protocols for diagnosing hydrocephalus Average organ doses measured from standard and acceptable reduced-dose protocols for diagnosing craniosynostosis Average organ doses measured from standard and acceptable reduced-dose protocols for diagnosing trauma follow-up Average organ doses measured from standard and acceptable reduced-dose protocols for diagnosing VP Shunt Distribution of exam indications for pediatric head CT at UF health

20 LIST OF ABBREVIATIONS 3D-CT AAPM ACR AEC AIDR AIDR-3D ALARA AP BEIR BMI CAP CMS CNR CSF CT CTDI CTDIvol DIR DLP EC FBP GM GWM HU Three-Dimensional Computed Tomography American Association of Physicists in Medicine American College of Radiology Automatic Exposure Control Adaptive Iterative Dose Reduction Adaptive Iterative Dose Reduction 3D As Low As Reasonable Achievable Anteroposterior Biological Effects of Ionizing Radiation Body Mass Index Chest-Abdomen-Pelvis Centers for Medicare & Medicaid Services Contrast-to-Noise Ratio Cerebrospinal Fluid Computed Tomography Computed Tomography Dose Index Volumetric Computed Tomography Dose Index Dose Index Registry Dose Length Product European Commission Filtered Backprojection Grey Matter Grey-White Matter Hounsfield Units 20

21 IC ICRP ICS IR IRB LCS LDCT LNT MDCT MTF NCRP NINDS NLST NPS OSL OSLD PACS PMMA PVC QC ROI SD SEMAR SNR SSDE Image Criteria International Commission on Radiation Protection Image Criteria Score Iterative Reconstruction Institutional Review Board Lung Cancer Screening Low Dose Computed Tomography Linear No Threshold Model Multi-Detector Computed Tomography Modulation Transfer Function National Council on Radiation Protection & Measurements National Institute of Neurological Disorders and Stroke National Lung Screening Trial Noise Power Spectrum Optically-Stimulated Luminescence Optically-Stimulated Luminescence Dosimeter Picture Archiving and Communication System Polymethylmethacrylate Polyvinyl chloride Quality Control Region of Interest Standard Deviation Single Energy Metal Artifact Reduction Signal-to-Noise Ratio Size-Specific Dose Estimates 21

22 TCM UF USPSTF VGA VGAS VP WL WM WW Tube Current Modulation University of Florida U.S. Preventive Services Task Force Visual Grading Analysis Visual Grading Analysis Score Ventriculoperitoneal Window Level White Matter Window Width 22

23 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy MEASURING ORGAN DOSES AND ASSESSING CLINICAL IMAGE QUALITY FOR THE PURPOSE OF COMPUTED TOMOGRAPHY PROTOCOL OPTIMIZATION By Izabella L. Lipnharski August 2017 Chair: Manuel Arreola Major: Biomedical Engineering Medical Physics The rapid technological development of CT imaging has resulted in a continuing expansion of CT examinations. Due to the radiation dose delivered by CT, it is important to ensure that the required diagnostic information is achieved with minimum dose to the patient. Methods were created to systematically assess clinical image quality within our radiology department. Various CT examination protocols, including pediatric body, pediatric head, adults with metal implants, and lung cancer screening, were investigated for potential dose reduction using eleven cadavers scanned on a clinical 320-detector CT scanner. Dose reduction was achieved by increasing the image noise acceptance level, as well as utilizing tube current modulation, iterative reconstruction, and organ shields. Subjective image quality was assessed using a task-based approach, scoring various clinically-relevant quality criteria on a three-point-scale to identify low-dose studies with acceptable image quality. Organ doses were directly measured to quantify organ dose savings achieved from modern technologies. Results showed that exam indications investigated in this work had room for dose reduction while maintaining diagnostic image quality. Furthermore, it was clear that diagnostic image quality varies 23

24 depending on the indication, and examination protocols should be optimized on an individual basis. 24

25 CHAPTER 1 INTRODUCTION 1.1 Growth of Computed Tomography Since its introduction in 1973, computed tomography (CT) has played a critical role in diagnostic radiology (1). Technological developments in CT imaging such as helical scanning and multiple-row detectors have allowed for image acquisition at faster speeds with lower levels of radiation. These developments have made CT the imaging modality of choice for a variety of indications and symptoms, increasing the number of CT exams performed. The number of CT studies in the United States increased about eight-fold since 1980, growing at about 10% per year (2). It is estimated that more than 62 million CT scans are performed per year in the United States, as compared with about 3 million in 1980 (3). The greatest increase in CT utilization has been in pediatric diagnosis, driven by the ability to perform the examination in a short period of time, and in screening asymptomatic adults for lung cancer, colon disease, and cardiac disease (4). 1.2 Increase of Radiation Exposure While CT imaging provides tremendous benefits in diagnostic medicine, it also utilizes higher levels of radiation than other imaging modalities. The National Council on Radiation Protection & Measurements (NCRP) Report Number 160, Ionizing Radiation Exposure of the Population of the United States (5), states that the per capita dose from medical imaging exams increased from 0.55 msv in 1980 to 3.0 msv in 2006, nearly a six-fold increase. The report also specifies that although CT accounts for 17% of all imaging procedures in the United States (6), it contributes to 49% of the total dose (5). 25

26 Due to the increase in usage and widespread attention from publications and media reports, the medical community has enforced an appropriateness criteria to ensure that CT examinations are only performed when absolutely necessary. Furthermore, when a CT examination is warranted, it is important to minimize the radiation exposure to as low as reasonably achievable (ALARA) while producing images of diagnostic quality. In order to have a clear understanding of appropriateness, it is essential that the public and medical community understand the clinical benefits and radiation risks associated with a CT examination. 1.3 Radiation Risks The biological effects of radiation exposure can be either stochastic or deterministic. Stochastic effects are random in nature. There is no threshold dose at above which stochastic effects definitely occur, and only the probability of the effects can be estimated. Stochastic effects can either be genetic or carcinogenic. However, there is no direct evidence that parents exposed to radiation cause additional genetic disease in their children, and cancer induction is the most important and feared effect. There has been documented evidence of radiation-induced cancer. The most thoroughly studied group for determining health effects of ionizing radiation were the Japanese survivors of the atomic bomb attacks on Hiroshima and Nagasaki in Developed by the National Research Council (NRC), the Biological Effects of Ionizing Radiation VII (BEIR VII) report describes the risks arising from exposure to low levels of ionizing radiation (7). The authors evaluated 93,000 survivors who were in Hiroshima or Nagasaki at the time of the bombings, and identified approximately 13,000 incidences of cancer and 10,000 cancer deaths. For those who received doses above 100 msv, excess cancers were observed, showing an increasing rate with increasing dose. To put 26

27 stochastic risks into perspective, the BEIR VII lifetime risk model estimates that if 100 people were exposed to a dose of 100 msv, approximately 1 person would be expected to develop cancer, while about 42 of those 100 people would be expected to develop cancer from other causes. Although statistical limitations made it difficult to evaluate cancer risk at doses below 100 msv, the committee concluded that risk continued in a linear fashion at lower doses without a threshold, and that there is a risk of a biological effect with any amount of radiation dose absorbed by an organ or tissue. This is where most of the controversy concerning radiation effects exists. The latency period between the time of exposure and the onset of cancer is 5 to 15 years for leukemia and 10 to 60 years for solid tumors. Therefore, it is difficult to prove that a cancer was caused by radiation exposure, as other contributors during the latency period may have been the true cause of cancer. This is especially true when the exposures are at low radiation levels such as those encountered in diagnostic imaging. The linear no-threshold model (LNT) is the accepted risk model describing the relation between low-dose exposure to ionizing radiation and the incidence of solid cancers that are induced by ionizing radiation. While stochastic risks are generally the main concern with diagnostic imaging procedures, deterministic effects are still possible. Deterministic effects require that the absorbed dose be above a known threshold and are proportional to the dose given. Although radiation doses from medical diagnostic procedures are usually far below the dose threshold, incorrect scan techniques or repeat CT examinations have caused for deterministic effects to arise in patients undergoing CT studies. Recently, a few hundred patients experienced skin erythema and hair loss following CT perfusion protocols using 27

28 incorrect scan techniques (8 10). Furthermore, in 2011, the International Commission on Radiological Protection (ICRP) revised its eye dose threshold for cataract induction from 2 Gy to 0.5 Gy for acute or protracted exposure.(11,12). Patients who undergo multiple CT studies may reach the cataract threshold, and therefore should have their lens dose monitored to prevent radiogenic cataract induction. 1.4 Need for Accurate Organ Doses In order to describe the actual risks arising from a CT exam, risk models must rely on accurate dosimetric measurements. Organ dose measurements provide the most accurate and dependable metric for determining risk from a CT exam. Furthermore, due to the concerns of the possible deterministic effects from radiological studies, and knowledge of identifiable threshold dose levels, it would be useful to include organ dose information in a patient s electronic medical record to reduce the likelihood of adverse effects. The methods for determining organ doses are varied, and some studies report doses that may be considered outdated or inaccurate. These problems can yield conclusions that are misleading and cause misconceptions about the severity of radiation risks from CT imaging. For example, a publication by Brenner and Hall estimated that up to 2% of all cancers could be attributable to the radiation from CT studies (4). However, the publication calculated lifetime attributable risks of cancer mortality from organ doses estimated in Britain in Those organ doses were from a different time-period of CT imaging prior to the implementation of helical scanning, multiple-detector imaging, and tube current modulation. Published organ doses need to not only be accurate, but they also need to take all clinical aspects of modern CT 28

29 imaging into account. With the fast speed at which CT technology is advancing, this is a challenging task to accomplish. 1.5 Protocol Optimization in Computed Tomography While technical advancements in CT scanners provide tremendous benefits in dose reduction and image quality, they also enhance the complexity of CT exams. Inherent differences in CT scanner design, manufacturer, and proprietary dose reducing methods can contribute to a variation in CT doses (13). Several studies have shown variation in CT radiation dose across protocols, scanners, and hospitals. One study evaluated radiation doses received by over 1,000 adult patients who were examined with CT studies across 4 hospital systems in California. For each type of CT examination, effective doses varied significantly within and across hospitals, with a mean 13-fold variation between the lowest and highest dose. Authors found that the effective dose for adult abdomen-pelvis CT studies ranged from 3 to 45 msv across the hospitals, and the mean effective dose for a stroke CT was 8 msv at one institution, and 29 msv at another. Another study found that for pediatric abdomen-pelvis protocols, children less than 5 years old received effective doses ranging from 3.2 to 30.2 msv, and children between ages of years old received effective dose ranging from 6.4 to 35 msv (14). The variation in dose across clinical sites reflects site-specific methods of choosing different technical parameters to answer the same clinical question. Thus, the dose an individual receives depends on the institution, the CT scanner, and the scan protocol. The wide variability of CT dose in clinical practice suggests an opportunity for reducing dose with protocol optimization. The optimization of CT protocols requires tailoring exam-specific scan protocols to patient age, patient size, region of imaging, 29

30 and clinical indication in order to ensure that the radiation dose delivered to each patient is kept ALARA for the clinical purpose of the CT examination. Diagnostic reference levels (DRL) were introduced by the ICRP in 1990 (15). DRLs are set at the 75 th percentile of national dose distributions based on wide-scale hospital surveys of mean doses for CT exams. They are used by hospitals to identify protocols that deliver unusually high patient dose and to determine whether patient protection has been properly optimized. Utilizing DRLs for the development of new protocols and monitoring of existing protocols is endorsed by several professional and regulatory organizations (15 19). Although DRLs promote awareness for improving patient radiation protection, they are not to be confused with the optimal dose for a particular procedure. Rather, they represent the dose trigger level at which dose appropriateness should be investigated for unnecessarily high doses. Once a CT exam is considered justified and appropriate for the clinical scenario, scan parameters should be optimized and dose reduction techniques used to perform the CT examination with the lowest possible dose (20 22). A qualified medical physicist should work with radiologists and technologists to determine whether the required level of image quality may be acquired with lower dose levels. However, reducing patient dose in CT is usually associated with a decrease in image quality, possibly compromising diagnosis. Therefore, the amount of dose reduction needs to be thoroughly assessed. Unfortunately, dose reduction assessment in patients is limited, since a stepwise assessment requires a large number of participants with the same indication, and repeated exams with different dose levels in one patient are prohibited. Hence, this 30

31 study evaluates dose reduction by scanning cadavers with different dose levels in order to identify the acceptable threshold of image quality with the minimum radiation exposure. 1.6 Specific Aims It was the goal of this research project to develop a systematic methodology for optimizing clinical protocols with cadavers using a twofold approach: assess clinical image quality with radiologists and quantify organ dose savings. To accomplish this goal, the following specific objectives were performed: 1. Develop a methodology for assessing clinical image quality for CT protocol optimization 2. Optimize CT protocols for patients with metal implants 3. Optimize CT protocols in lung cancer screening CT 4. Optimize CT protocols in pediatric body CT 5. Optimize CT protocols in pediatric head CT The remainder of this dissertation will provide a general background and describe in detail the methods used to achieve the goals listed above, the results obtained from this work, conclusions that can be drawn from this work, and suggestions for future improvements. 31

32 CHAPTER 2 PRINCIPLES OF CT TECHNOLOGY 2.1 A Brief History of CT The fundamental ideas on which CT was based on can be traced back to 1917 when the mathematician Johann Radon demonstrated that the distribution of a material in a three-dimensional object can be reconstructed from several two-dimensional projections of the object (23). In 1956, Bracewell applied Radon s mathematics to create an algorithm for reconstructing astronomical images (24). The first medical applications of these reconstructions were performed by the South Africa-born physicist Allan Cormack. In 1963, Cormack published a method of calculating radiation absorption distributions in the human body based on transmission measurements (25). The first application of this theory was by the British engineer Godfrey Hounsfield, who invented the first commercial CT scanner at EMI Central Research Laboratories and introduced the first patient CT examination in 1971 at the Atkinson Morley Hospital in London using a prototype brain CT scanner (1,26), offering substantial proof of the success of the method by detecting a cystic frontal lobe tumor. Figure 2-1 shows a picture of the EMI scanner and Figure 2-2 shows a picture of the first clinical image obtained. In 1979, Hounsfield and Cormack were awarded the Nobel Peace Prize for their significant contributions to medicine and science. 2.2 Development of CT Technology The revolutionary success of the initial CT scanner created enormous interest and led to an explosion of research and development by many groups. Since then, CT has made great technical advances in dramatically increasing the speed of scanning 32

33 and image reconstruction. Important milestones include the introduction of helical CT imaging in 1989 (27) and multi-detector row CT (MDCT) in 1998 (28) Helical Scanning In early CT scanners, electric power was transferred to the x-ray tube using high voltage cables. The gantry would make a full rotation in one direction, acquire an image slice, come to a complete stop, and then reverse back in the opposite direction to acquire a second image slice. This process resulted in long procedure times and poor temporal resolution. In the mid 1980 s, the power slip ring was introduced, which removed the need for cables, and transferred power from a stationary power source of electrical contactor brushes to a continuously rotating gantry without the need to start and stop. As the gantry continuously rotates, it acquires data as the patient is transported at a constant speed through the gantry. This concept, referred to as helical CT, allows for rapid scans of entire regions of interest, and became the standard of care for CT imaging by the mid 1990 s. Figure 2-3 illustrates the basic concept of helical CT Helical pitch Some aspects of helical CT are different from those of axial scanning. One example is how fast the table slides through the gantry relative to the gantry rotation time and the slice thickness acquired. This is called the helical pitch, and is defined as the table movement per rotation divided by the beam width. For a single-slice helical scanner, slice thickness and beam width are equivalent, and the pitch is given by: Pitch = TF SC (2-1) 33

34 where TF is the table feed in mm per 360-degree rotation, and SC is slice collimation in mm. For MDCT, the denominator factor in the pitch equation is replaced with the total thickness of all simultaneously acquired slices, or the nominal beam width, given by: Pitch = TF N SC (2-2) where TF is the table feed in mm per 360-degree rotation, N is the number of detector rows, and SC is slice collimation in mm. A pitch of 1 means that the x-ray beams from adjacent rotations are contiguous; pitches greater than 1 have gaps between the x-ray beams from adjacent rotations; and pitches less than 1 imply x-ray beam overlap Multiplanar imaging Another aspect associated with helical CT is slice interpolation. For axial CT, all data for slice reconstruction was collected within the slice before moving to the next position. However, in helical CT, the x, y, and z slice planes are equivalent, allowing for slices to be reconstructed at any position, called multiplanar reformation (MPR). Furthermore, the data can be reconstructed to provide three-dimensional CT images (3D CT), which are useful in cases where surgeons need to visualize complex anatomy to help them plan surgical procedures Multiple-Detector CT The main problem of helical CT was the high amount of heat deposited in the x- ray tube anode due to continuous scanning. A solution to the tube heating issue was to use the x-ray beam more efficiently. If the x-ray beam were widened in the z-direction, and if multiple detector rows were used to collect data from multiple slices, then the total number of rotations needed to cover the anatomy of interest could be reduced, and thus 34

35 the usage of the x-ray tube could also be reduced. This is the main idea of multidetector CT (MDCT). The main difference between the hardware of single-slice CT (SSCT) and MDCT is in the design of the detector arrays. In SSCT, the detector arrays consist of over 750 detector elements in a single row across the irradiated slice to intercept the x-ray fan beam. In the slice thickness direction, the single detector elements were long enough to cover the entire x-ray beam width. In MDCT, each of the individual long detectors are divided into smaller detector elements, as shown in Figure 2-4. In this way, MDCT allows for simultaneous acquisitions at multiple slices at different z-locations. In addition, in SSCT, the thickness of the x-ray beam affects the z-volume coverage and the z-axis resolution. In these cases, users were confronted with the requirements when large volume coverage and high z-axis resolution were necessary. In MDCT, while the x-ray beam still affects the z-volume coverage, the detector row collimation determines the z-axis resolution. Due to initial limitations of acquired and processing large amounts of data, the first MDCT scanners in 1998 were limited to 4 rows of detectors (28). As computing power advanced, more data channels were provided to create the 8-slice scanner. In 2002, 16-slice MDCT scanners were utilized with slices thinner than 1 mm. All vendors continued past 16-detector rows to wider coverage in the z-direction with thin image sections between 0.5 and mm. During 2003 and 2004, 32-slice and 40-slice scanners were being introduced. By 2005, 64-slice scanners were announced. In addition to the simultaneous acquisition of more slices, MDCT x-ray beam widths could 35

36 be widened. For example, 16-slice beam widths were up to 32 mm, 64-slice beams up to 40 mm wide, and now modern scanners with 320-slices are 160 mm wide. Together, with the efficient use of the x-ray tube, wider volume coverage, reduced scan time, and improved longitudinal spatial resolution, helical CT and MDCT have contributed to a remarkable enhancement of the clinical potential of CT. Many time-critical or resolution-critical applications, such as multi-phase dynamic studies, CT angiography of the heart and brain, virtual colonoscopy, and perfusion studies were made possible. 2.3 Dose Reducing Technologies Extensive research and development has been made to provide excellent image quality at the lowest possible radiation dose. The development of CT technology has created noteworthy systems that significantly improved its diagnostic performance. These innovations have not only increased the scope of CT applications but have also driven the need for new methods of reducing radiation dose. CT scanner manufacturers have contributed immensely to the ALARA initiative by producing technologies that reduce doses while maintaining image quality Tube Current Modulation Variations in attenuation throughout the patient s anatomy lead to differences in x-ray absorption and therefore image noise. By adjusting the tube current according to the patient s anatomy, image noise can be adjusted to maintain a constant level of image quality and to reduce the radiation dose to the patient. CT Tube Current Modulation (TCM) is a technical innovation that can reduce radiation dose using automatic exposure control (AEC) systems. 36

37 There are three general types of tube current modulation systems. In longitudinal (z-axis) TCM, the tube current is adjusted along the z-axis of the patient to provide uniform image noise as the attenuation varies among the different anatomic regions. Figure 2-5 illustrates the concept of longitudinal tube current modulation in a CT scan of a cadaver. Angular (x- and y-axis) TCM adjusts the tube current to normalize the amount of photons reaching the detector as the x-ray tube rotates about the patient from the anteroposterior direction to the lateral direction. As the x-ray tube rotates about the patient, the tube current can be varied according to the attenuation information on the CT projection radiograph, or in real time based on the previous attenuation measurement in the last 180 projection. The third type is called angular-longitudinal TCM. The combination of angular and longitudinal (x-, y-, and z-axis) tube current modulation involves variation of the tube current both during gantry rotation and along the z-axis of the patient, as illustrated in Figure 2-6. This is the most comprehensive approach to CT dose reduction because the x-ray dose is adjusted according to the patient-specific attenuation in all three planes. Tube current modulation systems are available from all major CT scanner manufacturers. Each manufacturer has a different name for their specific TCM system and method for defining the image quality in the user interface. GE s TCM algorithm, Smart ma, defines its Noise Index as the standard deviation of CT numbers within a region of interest in a water phantom. The patient s attenuation measured on the CT radiograph is mapped to the tube current values designed to maintain a constant image noise level as the attenuation changes during each gantry rotation. The user specifies the desired image quality, where the image noise is inversely proportional to the square 37

38 of the tube current. With Toshiba s TCM algorithm, Sure-Exposure 3D, users have to select a Standard Deviation (SD) of pixel values to specify the desired image quality. Like GE, the algorithm also compares the patient s CT radiograph to the standard deviation of a water phantom, aiming to match the image noise to the target standard deviation value. Siemens TCM system, CARE Dose 4D, makes use of a Quality Reference mas to define the effective mas, or the mas divided by the pitch, that would be required to produce a specific image quality in an 80 kg adult or 20 kg child for a given protocol. For specific patients, the tube current is based on the CT radiograph and fine-tuned by an online feedback system from the previous rotation to set the tube current according to the attenuation measured at each tube angle. Philips TCM algorithm, Dose Right, uses a reference patient image from a satisfactory exam with which image quality for future exams is matched. Rather than a noise index, this system expresses the required image quality in terms of an existing clinical image Therefore, it is difficult to compare different scanning protocols, since there is no measurable value related with the image quality in the reference clinical image Iterative Reconstruction The motivation for low-dose CT imaging and the shortcomings of FBP to provide acceptable image quality at low-doses fostered the rediscovery of iterative reconstruction (IR) algorithms as a promising tool to decrease radiation dose via noise reduction. IR was the initially proposed approach for data reconstruction in the early days of CT technology during the 1970s. However, due to its demanding mathematical computations and the large amount of data in CT imaging, IR was not practical for clinical purposes. With the exponential growth of computing power and data processing, 38

39 the utilization of IR methods has recently become a realistic option with reconstruction times acceptable for clinical workflow Filtered backprojection CT images have been reconstructed from raw data using filtered back projection (FBP) since the inception of the CT modality. FBP has the ability to generate CT images of acceptable image quality in a fast manner, and hence has been established in the clinical setting and utilized as the image reconstruction method of choice for over four decades. However, its inherently flawed assumptions regarding the physics of the scan acquisition ultimately results in noisy images. For example, the algorithm does not accurately model scanner geometry and optics, and fails to account for effects of beam hardening, scattered radiation, non-linear detector response, and noise statistics. As a result, images reconstructed with FBP can be affected with high levels of image noise, streak artifacts, or impaired low-contrast detectability in certain clinical scenarios such as those including obese patients, high density structures, or metallic implants Statistical iterative reconstruction Iterative reconstruction is superior to FBP because it makes more accurate assumptions about scanner geometry (including the dimensions of the focal spot, the size of each detector, and the size and shape of each image voxel) and incorporates system statistical information (including photon statistics and electronic noise) into the reconstruction process (29 32). Because noise can be suppressed to produce clearer images, not nearly as many photons are needed to generate a high-quality image of the object scanned. This presents for a remarkable opportunity for dose reduction with maintained, or even improved, image quality. Because the improved models are computationally more complex and are combined with multiple iterations of 39

40 reconstruction, IR methods demand longer reconstruction time than FBP reconstruction methods. Where FBP operates to solve a solution by projecting measured data backwards, IR uses a forward optimization oriented process (33). The algorithm starts with an estimate of the object scanned, typically a uniform matrix with attenuation values divided equally by the number of voxels in a row. A forward projection of the estimate creates artificial raw data, which are compared to the real measured raw data of the sonogram. The comparison is used to compute a correction term. The correction term is back projected onto the object. The algorithm builds on the estimate by cycling through iterations until the update for the current image estimate is considered small enough and the solution that most closely approximates the true object scanned is accepted as the final image. Statistical reconstruction methods can work in the raw data domain, in the image domain, or in the iterative reconstruction process (34). The benefit of working in either domain is that they are independent of the iterative reconstruction algorithm and are considered pre- and post-processing methods which use information based on a physical model. For example, in the iterative reconstruction process, noise is removed and the anatomical features of the scanned region are preserved or enhanced. Additional removal of noise is performed in image space Adaptive iterative reconstruction Similar to statistical iterative reconstruction, adaptive iterative reconstruction also incorporates a physical model of the CT system to accurately characterize the data acquisition process. This allows for substantial improvements in image quality, especially for low-dose CT scans, where the propagation of non-ideal data during the 40

41 image reconstruction becomes more significant than in routine CT scanning. The greatest benefit of adaptive iterative reconstruction is that it uses an FBP reconstruction as the first estimation to the iterative reconstruction algorithm. This approach helps shorten the longer reconstruction time of statistical iterative reconstruction, bringing the reconstruction time on the order of seconds, on par with FBP reconstructed scans. Adaptive statistical iterative methods have demonstrated a substantial opportunity for reducing radiation dose to patients. This is made possible through iterative noisereducing methods applied during acquisition, in the raw projection data and in the image space, as well as during post-processing. These techniques have integrated seamlessly into the clinic as reconstruction times remain relevant for clinical practice and image quality is superior to that achieved using FBP reconstruction methods. Considering that iterative reconstruction alone can result in patient dose reductions in the 25-80% range, it should be an important part of any overall CT radiation dose reduction program. Each department must identify the suitable tradeoff between patient radiation dose, image appearance, and CT exam diagnostic capability when incorporating iterative reconstruction over time. Each CT manufacturer offers different types of iterative reconstruction method with vendor-specific names and proprietary reconstruction methods: adaptive statistical iterative reconstruction (ASIR) and model based iterative reconstruction (MBIR) (GE Healthcare), adaptive iterative dose reduction (AIDR) (Toshiba Medical Systems), iterative reconstruction in image space (IRIS) (Siemens Medical Solutions), and idose (Philips Healthcare). 41

42 2.4 Organ Shielding Selective shielding of radiosensitive organs such as the eye lens, thyroid, breasts, and gonads has been used during CT examinations, especially in pediatric patients (35,36). Commercially available bismuth shields are placed over the surface of the organ so that the primary x-ray beam is attenuated before reaching the patient. The amount by which the xray is attenuated has been estimated by many studies. Some studies overestimated organ dose reductions by assuming the organ dose reductions were equivalent to the measured skin dose reductions, estimating dose savings of up to 67% (37,38). A study computed organ doses with Monte Carlo simulations, estimating organ dose reductions up to 30% (39). Despite the dose savings advantages of shields, concerns have been raised related to the impact on image quality. The use of shielding is generally not recommended, as the dose reduction can be achieved by decreasing x-ray tube current (40). However, global tube current reduction may not be easily implementable on all in a large multi-vendor hospital and as such, many facilities use bismuth shielding to easily and effectively reduce radiation exposure to the breast tissues. 42

43 Figure 2-1. First clinical prototype EMI brain scanner installed at Atkinson Morley s Hospital, London. Reproduced from G.N. Hounsfield, Computed Medical Imaging (26) with permission from The American Association for the Advancement of Science. Figure 2-2. First clinical image obtained from EMI CT scanner. Image was acquired in a woman with a suspected brain lesion, the scan clearly shows a dark circular cyst. Reproduced from G.N. Hounsfield, Computed Medical Imaging (26) with permission from The American Association for the Advancement of Science. 43

44 Figure 2-3. Illustration of principle of helical CT. The motorized table moves continuously through the gantry as the x-ray source and detectors rotate, producing a helical scan. Reproduced from DJ Brenner, Computed Tomography An Increasing Source of Radiation Exposure (4) with permission from the Massachusetts Medical Society. Figure 2-4. Illustration of single slice CT and multi-detector CT. Left image represents single, long elements along the z-axis, and right image represents MDCT arrays with several rows of smaller detectors. Reproduced from LW Goldman, Principles of CT: Multislice CT (41) with permission from the Society of Nuclear Medicine and Molecular Imaging. 44

45 Figure 2-5. Graph of tube current (ma) superimposed on a CT projection radiograph, with variation of the tube current along the z-axis, illustrating the concept of longitudinal dose modulation. Driven by attenuation data from the CT projection radiograph, the tube current increased at regions with increased attenuation, such as the pelvis, and decreased at regions with decreased attenuation, such as the lungs. Figure courtesy of author. Figure 2-6. Graph of tube current (ma) superimposed on a CT projection radiograph shows the variation in tube current as a function of time. Reproduced from CH McCollough et al, CT Dose Reduction and Dose Management Tools: Overview of Available Options (42) with permission from the Radiological Society of North America. 45

46 CHAPTER 3 DOSIMETRY IN COMPUTED TOMOGRAPHY In order to understand the several hardware and software features that have an impact on radiation dose to the patient, it is important to have a clear understanding of the various dosimetric quantities that characterize radiation exposure from CT scanners. The exposure geometry for CT is different from the standard radiographic x-ray system. In traditional projection radiography exams, the radiation dose is highest at the surface of the patient and reduces exponentially as it travels through the body of the patient. In CT, the x-ray tube exposes a narrow section of anatomy while making a full rotation around the patient, and continues with several rotations along the length of the patient. As a result, the dose in CT is more uniformly distributed throughout the patient. Because of its unique geometry, CT has its own set of parameters for radiation dose. 3.1 Multiple Scan Average Dose (MSAD) Early approximations of CT dose measured only the dose from a single scan acquisition. However, this method neglected the tails of the dose distribution caused by scattered radiation from adjacent scans, and therefore severely underestimated the dose delivered to a patient. In general, clinical CT examinations acquire images from multiple rotations of the x-ray source while the patient is translated through the gantry. The dose to the patient is the sum of the accumulated dose from all adjacent overlapped single scans. Provided that enough scans are acquired, the average dose over the central scan width will reach an equilibrium value, called the multiple scan average dose (MSAD) (43), as illustrated in Figure 3-1. The MSAD is the dose from a multiple scan examination, averaged over one scan interval in the central portion of the dose profile, as described in Equation 3-1: 46

47 MSAD = ( 1 + I I ) 2 D(z) dz I 2 (3-1) where MSAD is the average dose over a small interval from I/2 to I/2 about the center of the scan length for a scan interval I. D(z) is the dose at position z resulting from the series of CT scans. The SI units are in milligrays (mgy). The MSAD was measured with dosimeters that integrate dose over several minutes, such as film or thermoluminescent dosimeters (TLDs) by making multiple scans of a cylindrical phantom as it was stepped through the scan plane. These methods were laborintensive, had a heavy load on the x-ray tube, and involved tedious calibration, handling, and reading of the dosimeters. 3.2 CT Dose Index (CTDI) In 1981, Shope et al (44) provided a more practical method for estimating MSAD. The CT Dose Index (CTDI) was introduced, defined as the integral of the single scan radiation dose profile along the z axis, normalized to the thickness of the slice thickness. The general form of the CTDI is calculated as: CTDI = 1 + T D(z) dz (3-2) where D(z) is the radiation dose profile along the z-axis and T is the slice width. Therefore, it was presented that dividing the integrated absorbed dose by the beam width provides an estimate of the MSAD. Due to the speed and ease of CTDI measurements, which required only a single-scan acquisition, the use of MSAD declined, and CTDI became the standardized measurement technique. Several variations of CTDI exist, with unique measurements and calculations, and are described in the next sections. 47

48 3.2.1 CTDIFDA To standardize CTDI measurements, the Food and Drug Administration (FDA) introduced the integration limits of ±7T (45), where T represented the slice width. Furthermore, the total beam width was changed to include N, the number of acquired slices. As a result, the nominal beam width was represented by NT. The CTDI measurement made by using these limits is referred to as the CTDIFDA: CTDI FDA = ( 1 +7T NT ) D(z) dz 7T (3-3) To measure the dose profile over the exact limits of fourteen sections, measurements were conducted with film or a linear array of TLDs. Measurements could also be performed with a pencil ionization chamber, but its length of 100 mm meant that only 14 sections of 7-mm thickness could be measured. For thinner nominal sections, sometimes lead sleeves were incorporated to cover the areas of the ion chamber that exceeded fourteen section widths. The FDA also specified the scattering media for CTDI measurements. These included two polymethylmethacrylate (PMMA) cylinders with a length of 14 cm. To estimate doses for head examinations, a diameter of 16 cm is used, and for body examinations, a diameter of 32 cm is used. These phantoms are referred to as the head and body CTDI phantoms CTDI100 In practice, it is more convenient to measure CTDI over a fixed length of integration using a pencil ionization chamber with an active length of 100 mm. The CTDI100 allowed calculation of the index for 100 mm along the length of an entire pencil 48

49 ionization chamber, regardless of the nominal section width being used (43). This index is therefore defined as follows: CTDI 100 = ( 1 NT ) 5 cm 5 cm D(z) dz (3-4) CTDI100 is measured with a 100-mm-long, 3 cm 3 active volume pencil ionization chamber in the standard PMMA CTDI phantoms. Most commonly, CTDI100 is measured with a typical integrated exposure along the length of an ion chamber. This is equivalent to the following formula: CTDI 100 = ( 1 NT ) (f C E L) (3-5) where NT is the nominal beam width, f is the conversion factor from exposure to absorbed dose, C is the calibration factor for the electrometer, E is the measured exposure acquired from a single rotation with a beam profile of NT, and L is the active length of the pencil ionization chamber. Thus, the exposure measurement, performed with one axial scan in the PMMA phantoms for which CTDI is defined, results in a calculated dose index, CTDI100 (46). Figure 3-2 displays a photograph of equipment typically used to measure CTDI CTDIw The CTDI varies across the field of view. For example, the CTDI is usually a factor of two higher at the surface than at the center of the body phantom. CTDIw was introduced to provide a weighted average of the central and peripheral dose measurements (17,47). When calculated from measurements made in the standard phantoms, the CTDIw is weighted by one-third of the central position dose to two-thirds of the peripheral position dose. The definition is given by Equation 3-6: 49

50 CTDI w = ( 1 3 ) (CTDI 100 ) center + ( 2 3 ) (CTDI 100 ) periphery (3-6) CTDIvol The distribution along the patient z-axis is related to the spiral pitch in helical scanning, and presents challenges for identifying appropriate dose parameters to describe the radiation dose to the patient. In order to account for gaps or overlaps in between the radiation dose profiles from consecutive rotations, a correction factor is applied to the CTDIw to give the CTDIvol. The CTDIvol is the most widely used dose descriptor in CT, and is defined as: CTDI vol = CTDI w pitch (3-7) where pitch is defined as table distance traveled in one gantry rotation divided by the total collimated width of the x-ray beam. The measurement of CTDIvol is defined, and is required to be displayed on the scan console by the IEC standards on safety in CT. 3.3 Dose Length Product (DLP) While the CTDIvol is a measure of the absorbed dose at the central slice region of a 100-mm scanned volume, some consideration needs to be given for the physical length of the patient scanned. To better represent the total radiation dose imparted by a given CT examination, the CTDIvol can be combined with the scan length to compute the dose length product (DLP) (17): DLP = CTDI vol Scan Length (3-8) The scan length is measured in centimeters, and the DLP is given in units milligray-centimeters (mgy-cm). While the CTDIvol can only be used to compare the 50

51 absorbed dose for specific protocols, the DLP considers the integrated radiation output and can be used to determine the radiation risk attributable to the scan. 3.4 Effective Dose While CTDIvol and DLP provide information about physical measurements of dose, they do not describe the radiation risks associated with a CT examination. For this purpose, the effective dose (E) was introduced by the International Commission on Radiological Protection (ICRP) to reflect the risks of cancer induction from exposure to ionizing radiation (15). The effective dose, formerly called the effective dose equivalent, is expressed in units of millisieverts (msv) and defined as a measure of the risk of cancer induction in the patient from the effects of the radiation. It takes into account the absorbed dose received by specific tissues exposed to radiation and averages it to give a whole-body effective dose that would result in an equivalent stochastic risk. It is the weighted sum of the organ doses, described as: E = (ω T ω R D T,R ) T (3-9) where E is the effective dose, wt is the tissue weighting factor, wr is the radiationweighting factor, DT,R is the average absorbed dose to tissue T, T is the subscript for each radiosensitive tissue, and R is the subscript for each type of radiation, where in CT, only x-rays are present. The tissue weighting factors depend on the radiation sensitivities of the organs, and are described in Publication 60 of the International Commission on Radiological Protection (ICRP). Effective dose can be calculated with methods using anthropomorphic phantoms or computer simulations with Monte Carlo modeling. These methods are time consuming and require a high level of expertise. Alternatively, publications allow users 51

52 to approximate the effective dose for their own protocols by using the exam DLP and region-specific conversion factors, k (46,48). 3.5 Limitations of Dose Metrics Although CTDIvol and DLP are useful indicators of the radiation output, they are a standardized measure of radiation output, and are not to be confused for a direct measurement of patient dose. As a result, they have limitations in how they can be applied, and must be used with a clear understanding of the limitations. Effective dose is modeled using a standard patient, and the calculated risk is averaged over gender and age. Because each individual patient deviates from this single model, it is not appropriate to use effective dose for any one individual patient. It is intended for providing a broad measure of risk, and for comparing the relative risks of different diagnostic examinations. In addition, by comparing typical examination effective dose to background radiation dose from natural sources, patients are better able to put the risks associated with medical exposures into perspective. 52

53 Figure 3-1. Illustration of principles of MSAD. The radiation dose profiles from nine adjacent transverse CT scans along a line perpendicular to the transverse scans, when summed, produce the MSAD profile. Reproduced from J.A. Bauhs, CT Dosimetry: Comparison of Measurement Techniques and Devices (49) with permission from the Radiological Society of North America. Figure 3-2. Equipment typically used to measure CTDI100. Black arrow points to an electrometer, white arrow points to a 100-mm-long CTDI ionization chamber, and arrowhead points to a CTDI phantom made of PMMA. Reproduced from J.A. Bauhs, CT Dosimetry: Comparison of Measurement Techniques and Devices (49) with permission from the Radiological Society of North America. 53

54 CHAPTER 4 CT ORGAN DOSE MEASUREMENTS IN CADAVERS Accurate dose models that estimate effective dose rely heavily on the ability to approximate the dose to radiosensitive organs exposed during CT procedures. However, it is often difficult to determine the organ doses resulting from CT exams. Knowledge of patient-specific organ doses would provide a better estimate of radiation risk. It was the aim of this dissertation to report typical organ doses resulting from modern CT scanner technologies and to assess dose reduction achieved by protocol optimization. This work utilizes a reproducible method for measuring organ doses that was developed by Thomas Griglock at the University of Florida (50), described in the following sections. 4.1 Cadavers Utilized in this Work The use of cadavers for research of this type is exempt from review by the University of Florida Institutional Review Board (IRB). All subjects used in this study were fully intact embalmed adult cadavers obtained from the Anatomical Board of Florida located at the University of Florida in Gainesville, FL. It was decided to only acquire female postmortem subjects for the purpose of measuring the radiation dose delivered to radiosensitive breasts. Although the research group at UF Health has purchased 17 cadavers for dosimetry research, the work in this dissertation did not involve the first 8 cadavers. As a result, the cadavers utilized in this work, cadavers 9, 10, 12, 14, 15, 16, and 17, shall be referred to in this dissertation as Postmortem Subjects 1, 2, 3, 4, 5, 6, and 7 respectively. Furthermore, 4 cadavers were borrowed from the University of Florida s College of Medicine anatomy class to obtain lung cancer screening scans for image 54

55 quality analysis. These subjects shall be referred to as Postmortem Subjects 8, 9, 10, and 11. The height was measured from head to heel using a full-body CT scan and electronic calipers available on the PACS viewing software. The weight was obtained using a large animal scale. The height and weight were not measured for all cadavers due to some subjects being borrowed briefly for image quality analysis, or being returned to the Anatomic Board before the cadaver could be scanned from head to toe or weighed. The medical physics community has agreed that effective diameter is a better indicator of body composition and therefore should be used in place of body mass index (BMI) for investigating dose versus patient size trends (51). The effective diameter was measured on an axial image at the central slice of the scan range for all 11 cadavers using electronic calipers. Table 4-1 describes the weight, height, BMI, and effective diameter for the cadavers utilized in this work. 4.2 Organ Dose Measurements In order to accurately measure organ doses of a human undergoing a CT scan, a systematic measurement methodology was needed. Thomas Griglock found that using optically stimulated luminescent dosimeters (OSLDs) implanted into cadavers offered a suitable method for internal organ dosimetry (50) Optically Stimulated Luminescent Dosimeters Optically stimulated luminescent dosimeters (OSLDs) called nanodots (Landauer, Glenwood, IL) were used in this work due to their small size, linearity, efficiency, and practical use. Unlike thermoluminescent dosimeters (TLDs) that require a time-consuming heating and annealing process, OSLDs have a rapid readout with pulsed laser light and erasure with exposure to ultraviolet light for a short period of time. Furthermore, their small size with a length of 1 cm allows for easy access to internal 55

56 organs without significant disturbance of internal anatomy. The nanodots are composed of sensitive carbon-doped aluminum oxide material (Al2O3:C), allowing for accurate measurements at 2% (52,53) while spanning a wide dose range from 0.1 mgy to 3 Gy (53). After being exposed to radiation, nanodots were read using the MicroStar InLight reader (Landauer, Glenwood, IL). Figure 4-1 illustrates the nanodot, adapter, and the Microstar InLight reader used for immediate dose readout. The main concern of using OSLDs at diagnostic energy levels is the energy dependence. Studies have found correction factors for energy dependence for CT energy spectra and geometry (52,54). Although correction factors must be applied to the measurements, the advantage of small size and simplicity of readouts and erasure made nanodots the dosimeter of choice for postmortem organ dosimetry in CT Organ Dose Measurement Methodology The general concept behind the organ dosimetry methodology was to measure point doses in organs of interest in postmortem subjects. Due to inherent anatomical differences between postmortem subjects, the point dose measurement locations were chosen to represent an average organ dose. Only organs that were located within the scan range of each protocol were directly exposed to primary radiation from the x-ray source. Organs that were located outside of the scan range are assumed to only be exposed to a minimal amount of scattered radiation. All organ doses are reported in order to demonstrate the low levels of radiation received by organs not located in the scan range. A radiologist specialized in CT-guided biopsies, Dr. Sharat Bidari, inserted flexible polyvinyl-chloride (PVC) tubing into the organs of interest of each postmortem subject, with a sealed end located inside the subject, and an open end located 56

57 externally. Figure 4-2 shows a schematic of the location of the access site tubes within the lungs, breasts, liver, stomach, small intestine, ascending and descending colon, uterus, and ovaries. The OSLDs were placed within thinner plastic dosimeter holder tubes at precise intervals, maintaining the same depth in each organ across exams. The number of OSLDs per organ for each subject is described in Table 4-2. The dosimeter holders were inserted into the large tubes affixed in the organs of interest. The dosimeter holders were directly placed without the larger tubes within the left and right parts of the thyroid due to its small size and shallow depth. Also, for the smallest cadaver, Postmortem Subject 5, the breast tissue was too small to place larger tubes, therefore the dosimeter holders were placed directly within the breast tissue with a single dosimeter in each quadrant of the breast. After the larger tubes were inserted into the organs of interest, dosimeter holders with fiducial markers were placed at the locations of the dosimeter holders where OSLDs would be placed and inserted into the subject. Next, the subject was scanned and the radiologist confirmed correct nanodot positioning locations for organ dose measurements. In order to measure skin doses, OSLDs were arranged in a straight line, with 5 cm intervals, sealed with plastic to maintain the positions and protect the OSLDs from liquid. For the body scans, five OSLDs were placed on each skin strip, and three skin strips were placed on the postmortem subjects, distributed at the beginning, middle, and end of the scan range. The skin strips were placed from the midline of each subject and wrapped around the right periphery of the anatomy. For the head scans, ten OSLDs were placed on one skin strip, which was wrapped around the postmortem subject s forehead. 57

58 Before conducting any CT scans, the background doses of all OSLDs were read. After OSLDs were inserted within the subject, the localizer radiographs and scan protocols were conducted twice consecutively with the same dosimeters in the same positions in order to maintain dose differences less than 5% (50). After the CT scans were performed, the OSLDs were read with a MicroStar InLight reader. Readings were corrected for energy and scatter response using correction factors measured by Sinclair et al (54), and the doses were multiplied by the traditional f-factor to convert the dose to air to dose in tissue based on the effective energies utilized on the CT scanner. The average, maximum, and standard deviation of each organ dose was calculated. After the doses had been read, the OSL material from each nanodot was displaced from its plastic casing and exposed to UV light for at least 24 hours. If background doses were found to be over 1 mgy, longer erasure times were utilized to ensure proper erasure. 4.3 CT Scanner Utilized in this Work All CT scans were performed on a 320 detector-row CT scanner, the Toshiba Aquilion ONE, by an experienced CT technologist, Brian Cormack. The subjects were centered in the gantry and kept in the same position for all scans. All subjects were scanned in a supine position on the CT table, with their arms raised above the head for body scans, and with their arms at their sides for head scans, as per routine clinical practice Automatic Exposure Control Utilized in this Work Since patients come in all shapes and sizes, automatic exposure control systems have proven to be valuable in maintaining diagnostic image quality at a radiation dose suitable for each patient. Prior to each scan, frontal and lateral CT localizer radiographs, called scanograms, were acquired for the user to select the scan range. Toshiba s 58

59 automatic tube current modulation system, called Sure-Exposure 3D, utilizes the scanograms to automatically measure patient attenuation and adjust the tube current (ma) to maintain consistent image quality throughout the entire scan range (21,55). The tube current is controlled in the xy-plane and the z-axis, allowing for a more uniform image quality at a lower dose (56). The image quality of each slice is greatly influenced by the image noise. For this reason, Sure-Exposure 3D provides a noise parameter called the Target Noise Level, defined as the standard deviation (SD) of Hounsfield Units (HU) in a water phantom (57). The system prescribes a dynamic tube current that modulates during the gantry rotation based on desired image noise and slice attenuation. The system allows the user to set the desired target noise level as well as the range of minimum and maximum ma values within which the tube current can modulate. Since the target noise level is directly related to image noise, it is inversely related to the square root of dose. The majority of the dose reduction investigations conducted in this work reduce dose by selecting a higher target noise level, which in turn increases image noise and reduces dose Iterative Reconstruction Utilized in this Work The Toshiba Aquilion ONE CT scanner is equipped with an iterative reconstruction algorithm called Adaptive Iterative Dose Reduction 3D (AIDR-3D). The AIDR-3D algorithm uses a fully automated dose reduction technique that applies noise reduction based on the scan conditions and noise statistics and adapts to different anatomical regions, organs, and reconstruction kernels to ensure that maximum dose reduction and signal preservation is achieved. 59

60 AIDR-3D works in both the raw data and reconstruction domains. Once the scan is acquired with low-dose parameters, the number of x-ray photons detected is reduced, causing photon starvation in the projection data and image noise. AIDR-3D processing uses statistical and scanner noise models to reduce noise and artifacts in the raw data domain. The upper portion of Figure 4-3 illustrates the steps taken to reduce noise in the raw domain and produce a FBP image. An iterative technique is then performed in the reconstruction domain to simultaneously reduce noise, preserve sharp details, and smooth the image, with a sophisticated process optimized to the particular body region being scanned. Finally, as shown in the last step of Figure 4-3, a weighted blending is applied to the FBP reconstruction and the output of the iterative process to maintain the noise granularity. Unlike other vendors that allow the user to specify the percentage of blending, AIDR-3D s automated nature ensures the right amount of FBP and IR blending to preserve frequency characteristics and image texture. Sure Exposure 3D and AIDR 3D work in unison to reduce image noise, maintain spatial resolution, preserve structural edges, and allow substantial dose reduction on the order of 75% (58) CT Scan Protocols All of the clinical protocols utilized in this work have been standardized by the Radiology Practice Committee at UF Health to ensure uniformity across all scanners. One limitation of using embalmed postmortem subjects is the inability of injecting intravenous contrast. However, none of the protocols investigated in this work utilize contrast in the clinical examinations. The specific CT scan protocols are described in detail in Chapters 6 through 9. 60

61 Table 4-1. Height, weight, BMI, and effective diameters of cadavers utilized in this work Cadaver utilized at UF Postmortem Subject ID Height (inches) Weight (lbs) BMI (kg/m 2 ) Weight Category Effective Diameter (cm) 9 1 5'3" Overweight '3" Overweight Normal Obese Note: Subjects 3, 4, and 6 were not weighed, and therefore do not have weight, BMI, or weight category information. Subjects 4 and 6 did not receive a full body scan before returned to the Anatomical Board, and therefore do not have height information. Subjects 8 through 11 were borrowed temporarily within one day, and therefore were not weighed. 61

62 A B C Figure 4-1. Photographs of nanodot and adapter. A) Front of nanodot (Landauer, Glenwood, IL) case with serial number, B) back of nanodot with bar code and sensitive material displaced from case, and C) nanodot case slides in and out of the adapter for read-out in a microstar reader. Figure courtesy of author. Figure 4-2. Schematic of general access tube placement within the organs of interest 62

63 Table 4-2. Number of OSLDs utilized per organ for each Postmortem Subject Head Scans Subject ID Skin Brain Lens Thyroid Salivary Glands Torso Scans Skin Thyroid Breast Lung Liver Stomach SI Colon Ovary Uterus Note: Only Subjects 5 and 6 had craniotomies performed to measure doses in the head. Subject 6 was retrieved by the Anatomical Board before body dosimetry could be performed, and therefore only had dosimeters implanted in the head. 63

64 Figure 4-3. Schematic of the AIDR 3D reconstruction process. Adapted from Reference (58) 64

65 CHAPTER 5 ASSESSING CLINICAL IMAGE QUALITY FOR PROTOCOL OPTIMIZATION With CT technology expanding its applications to new areas, it is necessary to enforce the radiation protection principle to keep the radiation dose to patients as low as reasonably achievable (ALARA). At the same time, image quality must be maintained so that radiologists can answer the clinical question at hand. For this purpose, clinically relevant image quality parameters must be assessed to provide a clear understanding of the required image quality. This chapter describes the systematic methodology of assessing clinical image quality used in this dissertation for various CT exam indications. The findings from this work are implemented in the subsequent chapters to ensure that dose reduction does not come at the cost of impaired clinical image quality. 5.1 Image Quality in Computed Tomography Image quality is a broad term that can be interpreted in various ways depending on who is evaluating the image and for what purpose. There are many reasons for which image quality is assessed in a radiology department. For example, acceptance testing of new imaging systems verifies that the system meets specific performance criteria, and regular quality control testing of existing imaging systems ensures regulatory compliance and consistent system performance over time. Such regulatory tests are typically carried out by medical physicists using phantoms. In these cases, image quality takes the form of measurable image characteristics, both objective and subjective, and is based on the physics of how x-rays interact with matter and how the image reconstruction process produces images. Another task that requires image quality assessment is the clinical optimization of diagnostic protocols. Protocol optimization relies on the ability of a radiologist to make a 65

66 correct clinical diagnosis. This ability is task dependent, where the requirements for optimal image quality depend on many factors including the exam indication, the disease, the subtlety of radiological features, and the medical experience of the radiologist. Image quality in clinical CT examinations relies on both the scan acquisition settings and the image reconstruction techniques. The reconstruction kernel, slice thickness, and advanced noise-reducing algorithms have a strong influence on the final appearance of anatomical structures. As a result, image quality measurements of a phantom are not a true indication of the diagnostic performance, and as such, image quality assessments for protocol optimization must be carried out on clinical images (59). The difficulty in determining the optimal image quality for a particular CT examination has led to several methods of evaluating image quality. These methods involve quantitative physical measures as well as qualitative observer perceptions of clinical performance. This chapter discusses these types of image quality metrics and methods for evaluating them Measuring Quantitative Image Quality The Joint Commission requires that at least annually, a medical physicist conducts a performance evaluation of CT systems to ensure that the scanner positioning mechanisms, radiation dose, and image quality are consistent over time (60). Furthermore, the American College of Radiology (ACR) also requires that facilities accredited in CT conduct routine quality control (QC) tests (61). These two groups state that the physicist must measure and document the CT number accuracy and uniformity, high-contrast resolution, low-contrast resolution, contrast-to-noise ratio (CNR), image uniformity, and artifact evaluation. These objective parameters are measured annually 66

67 using CT images of contrast-detail phantoms such as the ACR Gammex 464 CT Accreditation Phantom. The following section will summarize quantitative image quality metrics CT number accuracy and uniformity The pixel values that compose a CT image are estimated from a set of x-ray measurements using a mathematical reconstruction process. The cross-sectional x-ray linear attenuation coefficients of the patient s anatomy are mapped to CT numbers, or Hounsfield Units (HU), representing the percent difference in linear attenuation coefficient between any materials and water, given by Equation 5-1. CT Number (HU) = 1000 μ p μ w μ w (5-1) In the equation above, μp and μw are the linear attenuation coefficients for a sample and for water, respectively. The CT number for water is zero, and for air is , since it has a negligible attenuation. Phantom manuals list predicted HU for the different inserts in the phantom. Variations in CT numbers of a scanner over time might indicate lack of consistency in calibration or possible changes in tube filtration High-contrast resolution Spatial resolution, or high contrast resolution, describes the CT system s ability to distinguish small objects closely spaced together on an image. In CT, spatial resolution is affected by the size of the detector, focal spot size, and the reconstruction algorithm. Several methods have been described to measure spatial resolution, such as scanning a line-pair phantom to visually identify the limiting resolution, or scanning a wire to calculate the modulation transfer function (MTF). An MTF curve characterizes the modulation factor as a function of spatial frequency. 67

68 Low-contrast resolution One of the main reasons that CT imaging has developed to an essential key of medical practice is due to its ability to detect small low contrast features, such as subtle tumors or lesions in soft tissues. The ability to detect a low contrast feature depends on the object s size, shape, contrast, noise, and the background material it is located in. The low contrast resolution of a system determines the system s ability to resolve low contrast objects of increasingly smaller sizes, and is typically measured with an observer subjectively detecting distinct objects in a phantom. Since image noise can affect the visibility subtle low contrast objects, it must be maintained low enough to allow for important objects to be differentiated from the background. CT image noise is the random variation of the HU values in each pixel, and is related to the number of x-rays reaching each detector. An increase in dose results in a decrease in noise, related to the inverse square root of the x-ray intensity. Exams acquired with low doses may suffer from high image noise that puts the patient at risk of an inaccurate diagnosis. However, an exam acquired with too high dose subjects the patient to unnecessary radiation. Therefore, noise and its effect on low contrast detectability is one of the most important characteristics associated with dose utilization. The variations of CT number can be estimated within a small area of a CT image of uniform material using a graphic region-of-interest (ROI) measurement to provide the standard deviation (SD) of that region. Although it is common to state CT noise as a standard deviation value, this practice can be misleading since CT image noise is not uniformly distributed as a function of spatial frequency. The image reconstruction process in CT imaging causes noise intensity changes as a function of spatial frequency. Therefore, standard deviation is not a very useful metric to compare dose 68

69 utilization of CT systems or to assume that it relates to low contrast detectability performance. The noise power spectrum (NPS) is a more complete description of the image noise, as it gives noise as a function of spatial frequency and accounts for noise with different textures. A noise power spectrum reflects the degree of randomness for every spatial frequency and the shape describes where the noise power is concentrated in the frequency space Evaluating Qualitative Image Quality As previously stated, the purpose of medical images is to provide sufficient information for a radiologist to diagnose the patient. Therefore, image quality should be measured by methods that address the effectiveness of an image to define anatomical structures which are relevant to make a diagnosis. Since very low-dose images are prone to producing high levels of image noise that interfere with image interpretation, the standard for image quality ultimately depends on radiologists preference. Therefore, effective methods of assessing image quality are needed for the proper application of ALARA and implementation of diagnostically acceptable optimized CT protocols. It is difficult to correlate objective image quality with the ability to make a clinical diagnosis (62). Studies in literature have attempted to correlate clinical radiologist image quality scores with physical measurements, but have agreed that radiological interpretation remains necessary for protocol optimization (62 64). The current consensus for defining diagnostic image quality is based on a taskbased approach (65), with the performance measured with the opinion of the radiologist relating their ability to perceive certain anatomical details or features in the image and their confidence in the perception of the details. The next sections discuss the 69

70 established methods for performing diagnostic performance tests that measure image quality with a set quality criteria (59,62,66) Receiver operator characteristic analysis The most widely used method for quantifying the diagnostic performance of clinical image quality is the receiver-operating characteristic (ROC) analysis (67). The ROC analysis measures how well a human observer can accurately detect the presence of an object. Using the responses, the observer s sensitivity and specificity can be calculated at several decision thresholds, describing the performance of an imaging system. Although ROC analysis is considered the gold standard of evaluating performance in medical imaging, the method does have some limitations. In order to produce statistically significant results, ROC methods require a large number of patient images with subtle pathology. It is also necessary to follow-up with the patients to have knowledge of their health state in order to categorize the image as normal or abnormal. Furthermore, because reader variation is the largest factor of the variance between two conditions being investigated, a large number of observers are required to read the images (68). For these reasons, ROC analysis may not be the ideal method for measuring clinical performance at a radiology department where the intention is to find the optimal dose level for a given examination Visual grading analysis A simpler approach for subjectively assessing image quality is to let observers score the visibility and reproduction of normal anatomical structures. In this way, image quality is determined by the visibility of anatomical structures, and is acquired in a relatively quick manner without the need for pathology. Using visual grading of the 70

71 reproduction of important anatomical structures in clinical images has become an established method for evaluating image quality, and is referred to as visual grading analysis (VGA) (59,62,66). VGA provides a simple method for quantifying subjective opinions and performing statistical analysis. The task for the observer is to decide whether the criteria related to the normal anatomy are fulfilled or not in an image. The evaluated image criteria can be selected so that they represent different characteristics that can be found for typical pathological structures, or can be selected from the criteria defined by the Commission of the European Communities (CEC) in the European Guidelines on Quality Criteria for Computed Tomography, which defines the performance level that is required to produce adequate image quality (17). VGA studies are considered valid if the anatomical structures are selected based on their clinical relevance and if the observers are experienced radiologists. As visual grading is based on images of normal patients, collecting a large number of cases is relatively straightforward. Furthermore, these types of studies are relatively easy to conduct, especially compared to ROC studies, which is important for optimizing protocols within a radiology department at the local level. In general, there exist two methods to visually grade the visibility of anatomical structures. Image criteria (IC) scoring requires the observer to assess the image and determine if the image criteria have been met or not in a yes-or-no fashion. An image criteria score (ICS) is then calculated as the as the ratio of the number of fulfilled criteria to the total number of criteria assessed (69). Visual grading analysis (VGA) uses a multi-step scale to improve the accuracy of the image quality assessment. The VGA method requires the observer to assess image quality based on the visualization and 71

72 reproduction of defined anatomical structures using a relative or absolute grading scale. The numerical scales are often expressed with words to facilitate interpretation and improve inter-observer agreement. In relative VGA, each image is compared with a reference image, where a scale of 0 implies visibility equal to the structure within the reference image, and a negative or positive value implies inferior or superior visibility, respectively. In absolute VGA, the observer gives a score about the visibility of each detail using an absolute grading scale, usually with three or five scale steps ranging from very bad to very good. The scale steps are converted to numerical values prior to analysis, where the lowest step may be number 1 and the highest step may be number 5. The scores are then used to calculate the visual grading analysis score (VGAS), which is the score averaged over all observers, cases, and criteria, shown in Equation 5-2. VGAS = O,I S C (5-2) N C N I N o where the score S C for image criteria C is related to observer O and image I, N C is the total number of image criteria, N I is the total number of images, and N o is the total number of observers (70). Differences in the ratings can be verified using the statistical analysis of variance (ANOVA) in combination with a test for multiple comparisons to reduce the risk of random significance. Both the IC and VGA methods have inherent mathematical flaws. In IC, there is no gradual transition from not acceptable to acceptable, and thus the observer may be challenged in decided whether the criteria is acceptable or not. Furthermore, no consideration is taken into how strongly the image was or was not acceptable, which may lead to challenges in interpreting the results. The VGA-score is often criticized 72

73 since the scale steps belong to an ordinal qualitative scale, and do not represent numbers on an interval scale, which is required for a true mean value (69). Furthermore, since the only information about score 2 and 3 is that 2 is less than 3, it is not suitable to add the ordinal steps. For this reason, the mathematical validity and statistical analysis of the VGAS-score is questionable. 5.2 Image Quality Evaluation Method Used in this Work It was the purpose of this dissertation to optimize diagnostic protocols by using a straightforward and practical approach that could be easily implemented on a local level within our radiology department. We first conducted a preliminary image quality study to evaluate overall image quality perception by asking radiologists to score the overall image quality of a CT scan. The radiologists responded that the score depended on what features they were assessing. The wide variety of anatomical regions, exam indications, and specific image quality features called for more specific image quality questions Task-Based Image Quality Analysis With the goal to focus on clinically relevant task-based image quality criteria, an absolute visual grading analysis (VGA) was used to evaluate the image quality of several CT exams and anatomic structures in this dissertation. VGA assumes that the level of visibility of anatomical and pathological structures are connected, so that if the visibility of normal anatomy is increased, then the visibility of pathological structures is also increased. As a result, we assumed that a VGA method would make it possible to predict the perception of structures within an image. 73

74 5.2.2 Ideal Grading Scale for Protocol Optimization The approach utilized to achieve diagnostic dose reduction at UF Health was to start with a standard scan protocol, gradually lower the dose by changing only one factor, ask the radiologists which images were acceptable, and identify the lowest dose scan that produced acceptable image quality. Studies often grade radiological image quality with a 5-point scale where a score of 3 describes average image quality, 1 and 2 represent inferior image quality (i.e. suboptimal, poor), and 4 and 5 represent superior image quality (i.e. good, excellent). A common issue of a 5-point scale is demonstrated in a study that assessed diagnostic acceptability on a 5 point scale (1: Unacceptable; 2: Suboptimal; 3: Average; 4: Good; 5: Excellent). Following radiological grading, the physicists deemed images acceptable if graded as 2 or above (71). Another study also used a 5-point scale (1: Extremely poor; 2: Poor; 3: Average; 4: Good; 5: Excellent) and later considered images acceptable if scored higher than a 2 (72). However, it was not clear whether the radiologists who graded images suboptimal or poor would welcome images of that quality. A fine 5-point scale may complicate the assessment of CT image quality in terms of diagnostic acceptability, whereas a coarser scale may do a better job at directly translating the radiologists perception of image quality acceptability. As a result, a 3-point grading scale was utilized for all image quality investigations in this work (1: Unacceptable; 2: Borderline Acceptable; 3: Acceptable). This scale removes the need for a subjective interpretation of the grading by the physicist, and requires the radiologist to select the limits of image quality acceptability, regardless of whether the acceptable image was adequate, good, or excellent. Radiologists were informed that a score of 1 meant the image was nondiagnostic and 74

75 they would request a rescan, 2 meant they could diagnose that particular study with reduced confidence but would not want to see images of that quality in the future, and 3 meant they were satisfied with the image quality and welcome additional images of that quality for interpretation Statistical Analysis An absolute VGA score (VGAS) was calculated for each reader using Equation 5-2. For studies that had at least three radiologists reading the images, the interobserver agreement for VGA scores was determined by calculating the intra-class correlation coefficient (ICC). The ICC ranges between 0 and 1.00, with values closer to 1.00 representing better reproducibility. ICC is interpreted as follows: An ICC of indicates slight reproducibility; an ICC of , fair reproducibility; an ICC of , moderate reproducibility; an ICC of , substantial reproducibility; and an ICC of greater than 0.80, almost perfect reproducibility (73). All calculations were performed with software (Excel Analysis ToolPak). For studies which received inter-observer agreement of at least moderate reproducibility, the VGAS was averaged over all image criteria and radiologists to form an average VGAS for each scan setting. For studies which received inter-observer agreement below moderate reproducibility, radiologist-specific VGAS were reported. For studies that had less than three radiologists reading the images, the inter-observer agreement was not determined and the radiologist-specific VGAS were reported. 5.3 Identifying Clinical CT Exam Indications for Protocol Optimization Although there exist a variety of reasons to perform a CT exam, protocol optimization relies on focusing on specific protocols, and when possible, indications for which the protocol would be utilized. For example, a chest CT exam may be ordered for 75

76 thoracic pathology, thoracic abnormalities, pulmonary disease, trauma to the chest, lung cancer, and more. Some of these indications may allow for lower dose, especially if they are conducted on at-risk populations. The first step in this aim was to identify which CT exam indications would be investigated for protocol optimization. As protocol optimization would be investigated with cadavers in this study, we were limited to protocols that did not use intravenous contrast. Furthermore, the cadavers presented with interesting anatomical features or body sizes, which was useful for investigating specific scan protocols. The initial CT scan of Subject 3 revealed that the subject presented with bilateral hip implants. With Toshiba s novel Single Energy Metal Artifact Reduction (SEMAR) algorithm available in the scanner, this was a fantastic opportunity to assess the image quality improvement with SEMAR, as well as to investigate SEMAR s usefulness at lower doses. As a result, Chapter 6 investigates CT scans of subjects with metal implants. Following the National Lung Screening Trial s findings (74), several institutions have implemented Lung Cancer Screening with low-dose CT (75). With more than 8 million Americans eligible for annual lung cancer screening (76), it was crucial to verify whether the dose can be lowered while producing images of diagnostic quality. Chapter 7 focuses on image quality assessment and organ dose measurements of Lung Cancer Screening CT. Owing to its very small size, Subject 5 resembled an average 10-year-old child based on its height and weight. With the great motivation of decreasing CT dose to pediatric patients, Subject 5 made it possible to measure typical organ doses and 76

77 assess image quality for low-dose pediatric chest and abdomen scans, described in Chapter 8. After meeting with several radiologists for image quality assessments for Aims 1 through 4, a pediatric neuro-radiologist requested assistance in investigating dose reduction to pediatric head CT scans. The radiologist addressed four indications that were currently being scanned with standard dose, but that produced unnecessary image quality for the purposes of assessing gross anatomy. As a result, dose and image quality assessments were conducted for pediatric CT scans for hydrocephalus, VP shunt, craniosynostosis, and trauma follow-up, shaping Chapter 9 of this dissertation. 5.4 Assessing the Usefulness of Postmortem Subjects in CT Image Quality Before utilizing human cadavers for CT image quality assessment, it was important to investigate whether they served as appropriate surrogates for living patients, and whether CT image quality analysis was possible in spite of anatomical differences between cadavers and patients Methods Radiologists specialized in specific radiological imaging fields were selected due to their expertise and familiarity with exam indications of interest. Dr. Troy Storey, Chief of Musculoskeletal Radiology, reviewed CT scans containing metal implants; Dr. Tan- Lucien Mohammed, Chief of Thoracic Radiology, reviewed lung cancer screening CT; Dr. Johnathan Williams, Chief of Pediatric Radiology, reviewed pediatric chest and abdomen CT; and Dr. Dhanashree Rajderkar, a pediatric neuroradiologist, reviewed pediatric head CT for hydrocephalus, VP shunt, trauma follow-up, and craniosynostosis. 77

78 Radiologists were asked to identify differences observed between a patient and a cadaver, and to explain whether the differences impaired their ability to score image quality. For indications that were not present in the cadaver, radiologists clarified whether they would be able to grade the image quality pertaining to that indication Results Dr. Storey stated that Subject 5 had high amounts of edema, making it challenging to visualize soft tissue planes and subcutaneous fat. However, it was possible to analyze bone details. Dr. Mohammed stated that the lung parenchyma appeared abnormal due to the lung being collapsed without ventilation. The postmortem subjects had limited separation of lung parenchyma, which appears different from human lungs. Although these differences affected the perception, the radiologist could still grade image quality by restricting the visualization to regions that appeared normal. Pulmonary edema was present with fluid infiltrating the surrounding lungs. However, central regions of the lungs were not affected by fluid infiltration, and were suitable regions for grading image quality. The subjects also presented with bilateral large pleural effusion and pulmonary edema, and modeled patients with interstitial lung disease, pulmonary metastasis, and pericardial effusions. Other findings not specific to lung parenchyma include the lack of vascular supply and blood flow, causing the airways to be filled with debris, the heart filled with air and congealing blood, decompressed vessels, and collapsed aortic vessels. Dr. Williams stated that Subject 5 presented with air in the ventricular chambers, fluid in the abdomen, and diffuse hyperdensity in the soft tissues, affecting the evaluation of soft tissues. The liver, spleen, and kidneys looked very similar to a human, but the pancreas appeared swollen with edema. 78

79 Dr. Rajderkar stated that the subject had reduced grey-white matter differentiation in the brain, a common effect of decomposition due to brain swelling (77), which affected image quality perceptibility, but still allowed relative image quality assessment based on the standard dose scan. Although some postmortem anatomy appeared different from humans due to postmortem and embalming effects, radiologists agreed that image quality was sufficient for investigating dose reduction, and that loss of soft tissue contrast and lung detail provided a worst-case scenario patient, where if they could diagnose the cadaver, they could definitely diagnose a patient with regular soft tissue contrast and lung detail. Apart from the initial questioning, thoracic radiologists stated they had difficulty in grading lung detail in Subject 4 and two musculoskeletal radiologists stated they were not able to assess soft tissue contrast in Subject 3. In these cases, image quality assessment was discontinued, and the observer scores were not recorded. 5.5 Creating Indication-Specific Image Quality Questionnaires The European Commission presented a list of image quality criteria as guidelines for CT examinations of adult patients.(17) However, the criteria are specific to body regions, rather than imaging indications. Furthermore, some anatomical structures and image features may be more important than others for the condition being investigated. With certain imaging indications requiring more image detail than others, it was crucial to investigate image quality specific to exam indications. Identifying the appropriate image quality demands a comprehensive understanding of the image quality requirements for each diagnostic task and how the scanning parameters affect image quality. This is a challenging task due to variations in radiologist preferences and scanner performance, requiring for physicists to work with 79

80 radiologists to identify optimal parameters for specific clinical exams. As a result, radiologists were utilized in this work to develop detailed task-oriented clinical image quality evaluations Methods Radiologists were presented with indication-specific postmortem CT images and image quality features listed in the European Commission s image quality criteria as well as those listed in radiologist diagnostic reports. The radiologists assessed each feature individually and identified useful measures of image quality for the specific scan indication. They also suggested additional features that would be useful for assessing image quality Results Because Subject 5 presented with loosening around the threads of the dynamic compression screw, as well as ossification, the musculoskeletal radiologist agreed that it was important to assess the confidence in diagnosing loosening and ossification, both common clinical features in orthopedic radiology. Because the streaking artifacts affected the bladder, the radiologist agreed that it was important to assess the confidence in diagnosing a bladder mass to evaluate visualization improvement with SEMAR. Furthermore, he stated a hematoma near the metal implants would be affected by the streak artifacts, and again selected the assessment of confidence in diagnosing a hematoma to assess improvement with SEMAR. Because streak artifacts were confined to the pelvis region nearby the prosthetics, the questions were asked in both the anatomical regions above the metal implants and in the regions containing metal implants. For CT scans with metal implants, the radiologists selected the questions streak artifacts, trabecular & cortical bone detail, confidence in diagnosing loosening, 80

81 confidence in diagnosing ossification, soft tissue contrast, soft tissue noise, confidence in diagnosing bladder mass, and confidence in diagnosing hematoma. The thoracic radiologist determined that the best clinical features for assessing lung cancer screening CT protocols were sharp reproduction of lung detail and diagnostic confidence in assessing nodules. He stated that although image noise, sharpness, and contrast are important questions, they are represented in the two clinical questions. Some radiologists may have different understanding of image sharpness, noise, and contrast, and focusing on clinical questions would better represent radiologists acceptance of image quality. Furthermore, the radiologist also identified different lung diseases among postmortem subjects, including bronchiectasis, cystic lung disease, and interlobular septal thickening. Although these findings are not characteristic of lung cancer screening, they are detailed features found in the postmortem subjects, useful for assessing loss of image detail at lower doses. As a result, the radiologist selected confidence in assessing lung disease as another suitable image quality feature. For lung cancer screening CT, the radiologists selected the questions sharp visualization of lung detail, confidence in diagnosing nodules, and confidence in diagnosing lung disease. The pediatric radiologist stated that the subject was a good candidate for assessing bronchiectasis, a symptom of common pediatric diseases such as cystic fibrosis. He also stated that chest wall injuries was a good parameter to assess bone injuries, trauma, sternal fractures, and rib fractures. Although there were some obvious fractured ribs with callus repair, it was important to focus the radiologists attention to new subtle fractures. The radiologist stated that the subject s kidneys were clearly 81

82 visible, thus were appropriate for assessing nephrolithiasis. However, it was not possible to assess urolithiasis, since the ureters and urinary bladder were not included in the scan range. Furthermore, the appendix was not visible due to the metal, ruling out assessment of appendicitis. He identified an umbilical hernia, and deemed it an adequate question. Liver and pancreatic lesions are usually detected with contrast and displacement of normal vessels, but this was not a good candidate for detecting small masses in the liver. He also explained that bone tumors are common in the pelvis, but not chest or abdomen. For pediatric chest CT, the radiologist selected the questions soft tissue contrast, visualization of lung parenchyma, confidence in diagnosing bronchiectasis, confidence in diagnosing chest wall trauma, and overall image quality acceptance for diagnosing chest CT. For pediatric abdomen CT, the radiologist selected the questions soft tissue contrast in large abdominal organs, noise in the liver, sharpness in the liver, confidence in diagnosing umbilical hernia, confidence in diagnosing kidney stones, and overall image quality acceptance for diagnosing abdomen CT The pediatric neuro-radiologist reviewed the image quality features on an indication basis. For hydrocephalus, it was important to visualize the size and shape of the ventricles. For assessing VP shunt follow-up, it was important to evaluate the size of the ventricular system and the position of the VP shunt catheter. The radiologist stated that even with the absence of a VP shunt placement, evaluating the ventricular system was sufficient to determine diagnostic confidence at lower doses. Radiological assessment of trauma follow-up CT images requires visualization of known lesions. For craniosynostosis, it was important to evaluate the shape of head. Because the four 82

83 indications in head CT focused on different image quality features, four separate observer studies were completed with different sets of images reconstructed to their ideal parameters, and clinical image quality was evaluated with indication-specific questions. For hydrocephalus, the radiologist selected the questions cerebrospinal fluid space over the brain, temporal horns of the lateral ventricle, frontal horns of lateral ventricle, third ventricle, periventricular interstitial edema, sylvian fissures, and confidence in diagnosing hydrocephalus. For VP shunt, the radiologist selected temporal horns of lateral ventricle, sharpness of ventricle outline, and confidence of diagnosing VP shunt complications. For trauma follow-up, the radiologist selected grey/white matter differentiation, white matter abnormalities, extra-axial fluid space, ventricular size, and confidence in diagnosing follow up trauma. For craniosynostosis, the radiologist selected shape of the head, sutures, cortical and trabecular bone, air filled compartments, and confidence of diagnosing craniosynostosis. 5.6 Conducting an Observer Study This last part of Aim 1 designs how the reading session should be conducted in a reproducible method to properly grade image quality while minimizing bias Methods The CT scanner has standard exam protocols available for use. For the majority of the indications, protocol selection was straightforward and readily available in the scanner. However, there were certain limitations in selecting some scan protocols. For example, ideally, protocol optimization should be investigated in a CT scanner that is clinically used for the indication being investigated. The musculoskeletal radiologists typically assess clinical orthopedic studies acquired on a Siemens CT scanner. However, because SEMAR reconstruction is only available on the Toshiba Aquilion 83

84 ONE CT scanner, postmortem investigations were carried out on the Toshiba Aquilion ONE CT scanner using a protocol called pelvis with hardware. Subject 3 was scanned with chest protocols in January 2015 in order to investigate potential lung cancer screening protocols that were introduced to UF Health in March The target noise level was increased to 20 SD in order to achieve lowdose scans, and a sharp lung point reconstruction algorithm was selected to provide thin slices for lung cancer screening. In order to ensure clinical relevance of the image quality assessment, we selected radiologists from relevant subspecialties. For the assessment of bilateral hip implants of Subject 3, 3 musculoskeletal radiologists, 1 thoracic radiologist, and 1 resident assessed image quality. For assessing lung cancer screening, 2 thoracic radiologists and 1 resident assessed image quality. For assessing pediatric body CT, 3 pediatric radiologists assessed image quality. In reading pediatric head CT, 1 pediatric neuro-radiologist and 1 general neuro-radiologist assessed image quality. Postmortem CT images were displayed at radiology workstations. Acquisition parameters were removed in order to blind radiologists to dose settings. In order to facilitate image comparison, multiple images were displayed across two monitors. For investigations involving several image series for comparison, matrices of 2 x 2 were selected on each screen, displaying a total of 8 images at a time. For investigations with fewer image series, or with more detailed comparisons, matrices of 1 x 2 were selected on each screen, displaying a total of 4 images at a time. In either case, the default image was displayed in the top left corner, and the remaining low-dose series were arranged randomly about the two screens so the readers were unaware as to how they 84

85 were acquired. Images were initially displayed at standard window and level settings utilized for the reconstruction and indication, but radiologists were permitted to change the window and level to their preferred settings. The radiologists were allowed to zoom in to allow the subject anatomy to take up the full extent of the viewing boxes. Before asking image quality features, radiologists were asked to identify differences between cadaver and human CT images for that specific scan indication. Based on these differences, they explained whether this affected their ability to grade image quality. Radiologists were reminded frequently throughout the reading session that they were not to grade based on postmortem differences. If necessary, radiologists were asked to scroll to another anatomical region with ideal viewing conditions. Image quality parameters were read out loud to the radiologist one at a time, and the radiologist assigned a score to each image series for that feature according to their individual perceptions of acceptability by using the three-point scale described in Section It was important to steer the grading process and remind the radiologist that just because the image appeared inferior to the standard, did not automatically make it borderline acceptable. We would clarify this by asking In spite of the inferior image quality, would you accept images of this quality in the future? If they responded that although it appeared inferior, but acceptable, then we would verify with them that it was an image of acceptable image quality Results After scanning cadaver images and conducting observer studies, it was clarified that different indications are ordered with different protocols, and radiologists regularly request to have images reconstructed to their preferred techniques either before or after the exam. Prior to scanning the subjects for indication-specific exams by selecting a 85

86 standard exam protocol, it is crucial to verify with the technologist and radiologist which specific protocol and reconstruction settings are utilized for the indication of interest. Musculoskeletal radiologists stated the standard scan of Subject 3 was acquired with too low of a tube voltage. Although we acquired with 135 kvp, their Siemens scanner acquires with 140 kvp. Since the Toshiba Aquilion ONE CT scanner has a maximum tube voltage of 135 kvp, dose was increased by reducing the noise target level from 12.5 SD to 10 SD in order to increase the effective mas from 25 to 50. They also found that the soft tissue images standard body axial at 1 mm x 1 mm were reconstructed too thin, adding noise to the slices, and the bone images were reconstructed too thick with bone sharp with boost kernel at 0.5 mm x 0.5 mm. The lead CT technologist at the orthopedic institute stated their CT scanner acquires at 140 kvp, and for patients containing metal implants, they typically provide bone images by reconstructing with a soft tissue algorithm and displaying with a bone window at the workstation. Specifically, images are reconstructed with a 2 mm x 2 mm soft tissue algorithm displayed in a soft tissue window, and 0.75 mm x 0.5 mm soft tissue algorithm displayed in a bone window. The recommended reconstruction parameters were utilized in future cadaver scans. When thoracic radiologists first saw the postmortem lung cancer screening CT images, they were acquired as a chest without contrast protocol with 120 kvp and 20 SD. The radiologists stated that the images were reconstructed correctly with 3 mm x 3 mm soft tissue images and 0.5 mm x 0.5 mm lung point images. However, upon investigating resultant doses and protocols recommended by AAPM, our group recommended to the lead CT technologist to create a novel lung cancer screening 86

87 protocol with minimum and maximum tube current values, as well as optimal reconstruction parameters described in Chapter 7. Pediatric radiologists stated that the image noise and contrast of the standard pediatric chest and abdomen scans matched that of a typical pediatric patient. This validated that for a postmortem subject weighing 87 lbs, it was appropriate to utilize the pediatric protocol less than 100 lbs with the standard reconstruction parameters. This was expected since scans were acquired on a scanner that the pediatric radiologists were accustomed to viewing images. The images were acquired using a protocol called Ped < 100 lbs chest at 100 kvp, 10 SD, reconstructed with a Ped Body Chest FC18 Soft tissue 3 mm x 3 mm algorithm and a Ped Lung Chest 3 mm x 3 mm algorithm. The abdomen images were acquired with a Ped < 100 lbs Abdomen at 100 kvp, 10 SD, and 3 mm x 3 mm Ped Abdomen algorithm. The lead CT technologist stated that although Subject 5 represents a typical 12 year old pediatric patient, all patients older than 8 years old are scanned with adult techniques, since the head size is essentially the same. As a result, all pediatric head scans in Subject 5 were acquired using adult techniques with 120 kvp and 270 ma. The pediatric neuro-radiologist stated that the scan acquisition parameters and reconstruction techniques were accurate for hydrocephalus at 3 mm, trauma follow-up at 3 mm, and craniosynostosis at 1 mm. CT-scans for follow-up of VP shunts had traditionally been acquired with standard non-contrasted head scans. However, starting in the Fall of 2015, these scans were acquired with a head CT protocol called stealth. Stealth CT exams are acquired without gantry tilt for pre-surgical planning, use a lower 87

88 fixed tube current of 250 ma, and a rotation time of 0.5 s. As a result, the CTDIvol is reduced from 70.8 mgy to 29.6 mgy, as compared to the standard head scan. 5.7 Discussion A limitation of this method is that image quality analysis was based on normal anatomical structure. Because it is difficult to obtain postmortem subjects with pathology, we were not able to have radiologists evaluate their confidence in the specific pathologies themselves. However, the radiologists stated that the questions asked in the image quality evaluation were a good alternative for evaluating anatomy lacking pathology. Furthermore, although acceptable image quality can be determined, the results are a function of the scan protocol, subjects scanned, scanner model, and radiologists involved. Therefore, results from protocol optimization research studies should be viewed as general recommendations and should be investigated with hospital-specific scanners and radiologists prior to clinical implementation. 88

89 CHAPTER 6 PROTOCOL OPTIMIZATION IN CT OF PATIENTS WITH METAL IMPLANTS CT imaging has become a routine tool in the assessment of patients with musculoskeletal disease. When evaluating orthopedic implants, CT can be used to assess the implant hardware, the interface between implant and bone, and the surrounding soft tissues to evaluate fractures, loosening, hematomas, tumors, or inflammation (78 80). Unfortunately, if metal objects are present in the region being scanned, the CT images will experience strong streaking artifacts that may impair the detectability of structures of interest. With the risk of image quality being affected by the presence of metal, we investigated methods for optimizing CT scan protocols to improve image quality for patients presenting with metal implants. 6.1 Presence of Metal in CT Imaging: A Retrospective Analysis at UF Health From 1997 to 2005, the number of joint replacement procedures performed annually in the United States grew by about thirty percent from 2.7 million to 3.5 million procedures (81). Specifically, the number of hip replacements increased by about thirtytwo percent from 290,700 to 383,500 procedures, with the majority of these performed on patients in their upper 60s (81,82). Owing to advances in orthopedic procedures and the aging of the Baby Boom generation who started reaching the age of 65 in 2011, the demand for joint replacement in the United States is projected to double in the next two decades (83). With the projected increase of hip replacement procedures, we can expect an increase in the number of patients with metal implants undergoing radiological assessment. In order to assess the prevalence of patients with metal implants examined with CT imaging at our hospital, a retrospective study was 89

90 conducted to document the presence and types of metal implants encountered in our patient population Methods With approval by the University of Florida Investigational Review Board (Study Number ), a log of 500 patients was generated by searching the radiology picture archiving and communication system (PACS) for any patient 50 years of age or older, male or female, who had a Chest-Abdomen-Pelvis (CAP) or Abdomen-Pelvis CT exam at UF Health between September 16, 2014 and September 16, The patients medical record numbers (MRNs) were used to locate the CT images correlating to the selected CT study. Once the CT images were located and reviewed using the PACS (Visage Imaging, Inc, San Diego, CA) on radiology department computers, all identifying information was removed, and patients were assigned a Subject Number from 1 to 300. It was noted whether the patient was male or female, the patient s age during the exam, the CT scan protocol, the CT scanner model, whether metal implants were present, and a description of the implant Results Of the 300 patients who were investigated, 152 were female (50.6%) and 148 were male (49.4%). The patient ages ranged from 50 to 102, with an average of 65.6 years and a median of 64 years. Furthermore, 229 patients had received Chest- Abdomen-Pelvis scans, and 71 patients had received Abdomen-Pelvis scans. Overall, out of the 300 patients, 82 patients (27%) had metal implants. Specifically, of the 229 Chest-Abdomen-Pelvis scans, 73 patients (31.8%) had metal implants, and of the 71 Abdomen-Pelvis scans, 9 patients (12.67%) had metal implants. Out of the 82 patients who had metal implants, 35 were female (42.6%) and 47 were male (57.4%). 90

91 Figure 6-1 shows the prevalence of metal implants found in the 300 patients, and Figure 6-2 shows the types and number of metal implants found in the 82 patients. The most common type of metal was fixation hardware in the sternum and spine (34 patients, 41%). Fixation hardware is typically implanted in the sternum following a median sternotomy for open heart surgery and in the spine to increase stability and improve bone union following complications including degenerative disk disease, infection, or trauma. Other common types of implants were cardiac grafts (6 patients, 7%) and pacemakers (6 patients, 7%. These types of metal implants caused some degrees of streaking, but it was not investigated whether they affected diagnostic confidence. The greatest amount of streak artifact, which completely obscured parts of the images, occurred in patients with metal hip implants. Seventeen patients (20%) presented with hip replacements, where four patients had a double hip replacement, seven patients had a left hip replacement, and six patients had a right hip replacement. Of the seventeen hip replacements observed, five were seen from the 73 Abdomen- Pelvis exams (6.8%) and twelve were seen from the 229 Chest-Abdomen-Pelvis exams (5.2%). 6.2 Effect of Metal Implants on Patient Dose Although studies have shown advantages of using tube current modulation techniques for reducing patient dose in CT exams of the abdomen or pelvis, the effects of metallic prosthesis on tube current modulation methods remain undocumented. If a CT scanner s automatic tube current modulation system detects increased attenuation at a region containing metal implants, it will deliver a higher amount of tube current in that region to ensure photon transmission through the patient. With the potential for 91

92 increased radiation exposure to the patient, this study investigated the effect of orthopedic metallic prosthesis on patient dose Methods Three similar sized postmortem subjects, Subjects 1, 2, and 3, were acquired for this investigation. Subjects 1 and 2 did not present with metal implants, whereas Subject 3 had metal sternal fusions, lumbar fusions, and bilateral hip replacements. The effective diameter of each subject was calculated by measuring the anteroposterior (AP) and lateral (LAT) dimensions at the central slice of the scan (51). Subjects 1, 2, and 3 had effective diameters of cm, cm, and cm, respectively. The three subjects were scanned with a standard clinical adult CAP protocol with scan techniques listed in Table 6-1. The tube current was adjusted automatically by the scanner s Sure Exposure 3D algorithm and dose reduction was achieved via an iterative reconstruction algorithm, AIDR-3D Results As shown in Table 6-1, Subjects 1 and 2 had similar scanner-reported radiation output values, with percent differences of 11%, 3%, and 1% for effective mas, CTDIvol, and DLP, respectively. Subject 3 had higher values compared to Subjects 1 and 2, with percent differences of 56%, 32%, and 37% for effective mas, CTDIvol, and DLP, respectively. Figure 6-3 displays the scanogram image of Subject 3 with the tube current plots of all three subjects superimposed for comparison, showing that the tube current increased in Subject 3 at the regions containing metal. The tube current also increased at the regions of the arms of Subject 3, since a broken bone in the shoulder did not permit to completely remove the arm above the head, and regions of the upper arms were at the side of the head for Subject 3. The increase in tube current at the 92

93 region of the arms also contributed to an increase of the effective mas, CTDIvol, and DLP. However, the tube current plot still shows the prominent rise in tube current at the regions of metal. Table 6-2 lists measured organ doses for the three subjects, reported as average organ dose, maximum organ dose, and standard deviation. Figure 6-4 displays the average organ doses measured for the three subjects. Subjects 1 and 2 had a maximum organ dose difference of 1.9 mgy in the stomach. Subject 3 had larger dose differences of 3.0 mgy in the stomach, 3.2 mgy in the small intestine, 2.5 mgy in the colon, and 2.3 mgy in the ovaries, relative to a subject without metal implants. With this small relative increase in patient dose, the main issue of metal implants is perhaps not in patient dose, but in image quality, as described in the next section. 6.3 Effect of Metal Implants on Image Quality Studies have shown that the presence of metal objects in the scan field of view creates metal artifacts that impair the evaluation of bone and soft tissues nearby the implant, causing many exams to be considered uninterpretable (80,84). The metal artifacts arise from x-rays being highly attenuated by the metal, leading to photon starvation, noise, beam hardening, scattering, and nonlinear partial volume effects. This called for an investigation to evaluate the extent to which metal implants affect image quality for cadavers with metal implants scanned on a Toshiba Aquilion ONE CT scanner Methods Scanning with a CAP protocol Subject 3 was scanned with a standard clinical adult CAP protocol with scan techniques listed in Table 6-3. Images were reconstructed with a soft tissue algorithm 93

94 with axial slice thickness of 5-mm at 5-mm intervals and shown to 7 experienced radiologists and 4 radiology residents described in Table 6-4. Radiologists graded the lung detail in the chest, soft tissue visualization in the abdomen, and soft tissue visualization in the pelvis on a 3-point scale. VGAS and inter-observer agreement were calculated using methods described in section Scanning with a pelvis protocol Following the general investigation with a CAP scan, we investigated what scan parameters were modified in a clinical scan for a patient presenting with known metal implants. A CT scan protocol called Pelvis-With-Hardware (PWH) was included in the scanner to be utilized on patients with major metal implants. This protocol used a volumetric scan mode and a higher tube potential of 135 kvp in order to minimize metal artifacts. As a result, Subject 3, who had bilateral hip prosthesis, and Subject 5, who had a bilateral intramedullary rod with dynamic compression screws, were scanned using the recommended clinical PWH CT scan protocol. Subject 3 was scanned with a clinical PWH scan protocol with scan techniques listed in Table 6-5. Images were reconstructed to 1-mm thick soft tissue slices with 1- mm intervals and 0.5-mm bone slices with 0.5-mm intervals and shown to 5 specialized radiologists described in Table 6-6. The radiologists graded soft tissue visualization and bone detail in regions above the metal implants and directly nearby the metal implants on a 3-point scale. Subject 5 was scanned with a clinical PWH scan protocol with scan techniques listed in Table 6-7. Reconstruction settings differed from those used in Subject 3. While edge-enhancing filters such as the bone algorithm is generally preferred to display fine bone detail, it can accentuate the appearance of metal-related artifacts (80). Smooth 94

95 reconstruction filters such as the soft tissue algorithm can help to reduce metal artifacts, at the expense of reduced spatial resolution (78,80,85). Furthermore, thin-section acquisitions help to minimize artifact by reducing partial volume averaging, while thicker reconstruction helps to minimize the effect of artifacts by better averaging of the signal within the voxel and by increasing the available signal-to-noise ratio. In order to improve image quality, the investigation with Subject 5 was reconstructed with 2-mm thick axial slices with 2-mm intervals with soft tissue reconstruction algorithm and 0.75-mm thick axial slices with 0.5-mm interval with soft tissue reconstruction algorithms. The thicker 2-mm slices were displayed in the soft tissue window setting, and the thinner 0.75-mm slices were displayed in the bone window setting. The musculoskeletal radiologists detected bone loosening around the threads of the dynamic compression screw component in femoral head and neck of Subject 5. They also detected ossification, or bone formation at an abnormal anatomical site. Based on these findings in the subject, additional questions were added to the image quality assessment of Subject 5. Three musculoskeletal radiologists described in Table 6-8 evaluated bone detail, confidence in grading bone loosening, confidence in grading ossification, image artifacts, soft tissue contrast, and soft tissue noise in superior pelvic regions not containing metal implants and inferior pelvic regions containing metal implants. VGAS and inter-observer agreement were calculated using methods described in section Results Scanning with a CAP protocol Image quality scores for the CAP CT scan of Subject 3 are listed in Table 6-9. All radiologists graded the lung detail in the chest and the soft tissue visualization in the 95

96 abdomen to be of acceptable image quality. However, all eleven radiologists found that the streak artifacts impaired visualization of intra-abdominal contents, muscle detail and soft tissue interfaces in the pelvis, scoring the image quality in the pelvis to be of unacceptable image quality. Perfect inter-observer agreement among the participating radiologists was found by means of an intra-class correlation coefficient of 1.00 (p = 0) for the VGAS. Since all radiologists graded each feature with the exact same score, Figure 6-5 displays the region-specific VGAS averaged across all readers Scanning with a pelvis protocol Image quality scores for the PWH CT scan of Subject 3 are listed in Table Nearly perfect inter-observer agreement among the participating radiologists was found by means of an intra-class correlation coefficient of 0.89 (p < 0.001) for the VGAS. As a result, image quality scores were averaged over all readers and displayed as VGAS in Figure 6-6. In regions above the metal implants, soft tissue visualization and bone detail were scored as acceptable by all five radiologists. In regions nearby the metal implants, the scores decreased to borderline or unacceptable by all five radiologists. For example, soft tissue visualization was graded as unacceptable by the thoracic radiologist, resident, and one musculoskeletal radiologist. Two musculoskeletal radiologists graded the soft tissue visualization as borderline acceptable. The bone detail was graded borderline acceptable by all five radiologists. Image quality scores for the PWH CT scan of Subject 5 are listed in Table Substantial inter-observer agreement among the radiologists was found by means of an inter-class correlation coefficient of 0.78 (p < 0.001) for the VGAS. Therefore, image quality scores were averaged over all readers and displayed as VGAS in Figure 6-7. The VGAS for image artifacts and soft tissue contrast decreased in regions with metal 96

97 implants. The bone detail received a VGAS of borderline acceptable for regions with and without metal since the smooth kernel reconstruction compromised spatial resolution and therefore bone detail. 6.4 Effect of a Metal Artifact Reduction Algorithm on Image Quality Despite recent developments in CT technology, overcoming the inclusion of metal artifacts in CT still presents a challenge. Since the early 1980s, several methods have been developed for reducing metal artifacts, including increasing the tube current or tube potential, using dual-energy acquisition, and using post-processing reconstruction algorithms (78,80). Although these techniques have been somewhat successful in reducing metal artifacts, the clinical use of these techniques is limited. Increasing the tube current or tube voltage results in higher radiation dose, and although it may reduce the noise in the projection data, it does not correct for other data inconsistencies caused by metal implants such as photon starvation and nonlinear partial volume averaging (86). Therefore, the streaking and shading artifacts cannot be avoided even though a higher radiation dose is delivered to the patient. Although dual energy acquisition systems have proven useful in reducing streak artifacts with post processing, this technique requires a CT system capable of dual-energy acquisition (87,88). In recent years, Toshiba released its innovative raw data-based iterative reconstruction algorithm called Single Energy Metal Artifact Reduction (SEMAR) (89,90)s. SEMAR employs a sophisticated reconstruction algorithm that reduces metal artifacts by extracting the metal prior to reconstructing the image, and adding the metal back in the image domain at the end. The following section describes an investigation 97

98 into SEMAR s usefulness in the clinic for the assessment of bone and soft tissue in postmortem subjects containing metal implants Methods Scan techniques for subject 3 With SEMAR reconstruction currently only available for volumetric scans, Subject 3 was scanned with a PWH scan protocol with techniques listed in Table 6-5. The raw image data was utilized to reconstruct image sets without SEMAR and with SEMAR applied. Images were reconstructed to 1-mm thick axial slices with 1-mm intervals with soft tissue reconstruction algorithm and 0.5-mm thick axial slices with 0.5-mm intervals with bone reconstruction Scan techniques for subject 5 Subject 5 was scanned with a PWH scan protocol with techniques listed in Table 6-7. Images were reconstructed with and without SEMAR. Images were reconstructed to 2-mm thick axial slices with 2-mm intervals with soft tissue reconstruction algorithm, shown in soft tissue window, and 0.75-mm thick axial slices with 0.5-mm interval with soft tissue reconstruction algorithm displayed in a bone window Image quality evaluation For Subject 3, the radiologists were asked to assess soft tissue visualization and bone detail in the regions containing metal with and without SEMAR reconstruction using a 3-point scale. For Subject 5, the radiologists were asked to grade bone detail, bone loosening, soft tissue contrast, soft tissue noise, image artifacts, confidence in assessing bladder mass, confidence in assessing hematoma, and ossification in regions containing metal with and without SEMAR reconstruction using a 3-point scale. VGAS 98

99 and inter-observer agreement were calculated using methods described in section Results Soft tissue and bone reconstructed images with and without SEMAR are displayed in Figure 6-8 for Subject 3 and Figure 6-9 for Subject 5. Because SEMAR is only applied in the reconstruction setting, the dose values remain unchanged whether SEMAR is utilized or not Image quality scores for subject 3 Image quality scores for Subject 3 scanned with a PWH CT protocol are displayed in Table Fairly reproducible inter-observer agreement among the participating radiologists was found by means of an intra-class correlation coefficient of 0.06 (p > 0.001) for the VGAS. Because the inter-observer agreement was below moderately reproducible, image quality scores are displayed for each individual radiologist in Figure For all five radiologists, image quality scores for soft tissue visualization stayed the same or improved with SEMAR reconstruction. The radiologists stated that the use of SEMAR improved image quality and diagnostic confidence for the analysis of the periarticular soft tissue structures of the hip. Furthermore, pelvic organs that were often completely obscured by the metallic artifacts became identifiable. In analyzing the bone detail in the presence of metal, without SEMAR reconstruction, radiologists graded the bone detail with borderline acceptable and acceptable image quality. With SEMAR reconstruction, the image quality scores for bone detail varied. The first musculoskeletal radiologists stated that the images reconstructed with SEMAR had higher image noise and simulated a low-dose image. He confirmed that the image without SEMAR had higher diagnostic quality than the 99

100 image reconstructed with SEMAR. His image quality score reduced from a 3 to a 1 when SEMAR was enabled. The second musculoskeletal radiologist stated that the images reconstructed with SEMAR displayed a loss of image detail, making it possible to miss a fracture. He also preferred the image without SEMAR, despite the streak artifacts. His image quality score stayed the same as a 2 for both reconstructions. The third musculoskeletal radiologist stated that the image reconstructed with SEMAR made him question his diagnosis, and he preferred the image without SEMAR. He said he could minimize the streak artifacts by changing the window and level settings, but SEMAR introduced new artifacts into the image that he could not reduce. His image quality score stayed the same as a 2 for both reconstructions. Radiologists 4 and 5 were not specialized musculoskeletal radiologists, so although one graded the bone detail to improve with SEMAR, it might be due to the fact that the specific radiologist was not accustomed to reading patient images for diseases involving bone detail Image quality scores for subject 5 Image quality scores of Subject 5 scanned with a PWH CT protocol are displayed in Table Moderate reproducible inter-observer agreement among the participating radiologists was found by means of an intra-class correlation coefficient of 0.52 (p < 0.001) for the VGAS. Therefore, image quality scores were averaged over all readers and displayed as VGAS in Figure The VGAS for bone detail, image artifacts, soft tissue contrast, confidence in diagnosing a bladder mass, and confidence in diagnosing a hematoma improved when SEMAR was utilized. The VGAS score for confidence in grading loosening, confidence in grading ossification, and soft tissue noise stayed the same when SEMAR was utilized. However, these three criteria received a score of 3 with and without SEMAR. Therefore, even if SEMAR improved the quality, the score 100

101 could not have been improved any further. The radiologists stated that the bone detail was not as affected with the new reconstruction technique utilized in Subject SEMAR s Performance with Varied kvp and ma It is of interest to determine whether SEMAR is improved with higher dose acquisitions, and whether it remains useful at low dose acquisitions Methods Scan techniques for subject 3 CT scans with SEMAR reconstruction were acquired with PWH protocols with altered acquisition parameters listed in Table 6-14 for Subject 3. In Table 6-14, Scan 2 represents a standard scan that serves as a reference for the low and high dose protocols; Scan 1 represents a protocol with increased dose, achieved by demanding less image noise and reducing the target noise index from 12.5 SD to 10 SD; Scans 3 through 7 represent protocols of reduced dose, achieved by reducing the voltage from 135 kvp to 120 kvp, and by increasing the target noise index from 10 SD up to 20 SD. Images were reconstructed with a soft tissue algorithm with 5-mm thick axial slices with 5-mm intervals and with a bone algorithm using 1-mm thick axial slices with 1-mm intervals, resulting in a total of fourteen image series of varied kvp, noise index, and reconstruction algorithms Scan techniques for subject 5 CT scans with SEMAR reconstruction were acquired with PWH protocols with altered acquisition parameters listed in Table 6-15 for Subject 5. In Table 6-15, Scan 3 represents a standard scan; Scans 1 and 2 represent protocols with increased dose with reduced target noise indices; and Scans 4 through 6 represent protocols with decreased dose with increased target noise indices. The images were reconstructed to 101

102 2-mm thick axial slices with 2-mm intervals with soft tissue reconstruction algorithm and 0.75-mm thick axial slices with 0.5-mm interval with soft tissue reconstruction algorithms. The thicker 2-mm slices were displayed in the soft tissue window setting, and the thinner 0.75-mm slices were displayed in the bone window setting and shown to three musculoskeletal radiologists Image quality evaluation For Subject 3, three musculoskeletal radiologists, one thoracic radiologist, and one radiology resident described in Table 6-6 assessed 14 different image series with and without SEMAR, grading soft tissue visualization and bone detail in the regions containing metal using a 3-point scale. For Subject 5, Radiologists described in Table 6-8 were asked to grade bone detail, loosening, soft tissue contrast, soft tissue noise, image artifacts in pelvic regions not containing metal implants, and bone detail, loosening, soft tissue contrast, soft tissue noise, confidence in assessing bladder mass, confidence in assessing hematoma, image artifacts, ossification in the regions containing metal using a 3-point scale Results Image quality scores for subject 3 Image quality scores for soft tissue visualization and bone detail of Subject 3 are summarized in Table Fairly reproducible inter-observer agreement among the participating radiologists was found by means of an intra-class correlation coefficient of 0.07 (p > 0.001) for the VGAS. As a result, the image quality scores are displayed for each radiologist in Figure 6-12 for soft tissue visualization and Figure 6-13 for bone detail. 102

103 For the standard scan (protocol 2), the soft tissue visualization stayed the same (20%) or increased (80%) when SEMAR was utilized. For low-dose scans (protocols 3 through 7), the soft tissue visualization stayed the same (48%) or increased (52%) when SEMAR was utilized. For the high-dose scan (protocol 1), the soft tissue visualization stayed the same (40%) or increased (60%) when SEMAR was utilized. Overall, the soft tissue visualization was the same (40%) or better (60%) with SEMAR reconstruction in all of the cases. In the 14 cases where the image quality score stayed the same, 4 of them received a score of 3 with and without SEMAR. Therefore, even if SEMAR improved the quality, the score could not have been improved any further. As a result, SEMAR improved soft tissue visualization 60% of the time, did not improve image quality 28.6% of the time, and was unclear (with a maximum image score) 11.4% of the time. For the standard scan (protocol 2), the bone detail stayed the same (40%), decreased (40%), and increased (20%) when SEMAR was utilized. For low-dose scans (protocols 3 through 7), the bone detail stayed the same (48%), decreased (48%), and increased (4%) when SEMAR was utilized. For the high-dose scan (protocol 1), the bone detail stayed the same (60%), decreased (20%), and increased (20%) when SEMAR was utilized. Overall, the small frequency of bone detail improvement (8.6%) suggests SEMAR is not beneficial for bone detail reconstructed with bone algorithms Image quality scores for subject 5 Image quality scores for Subject 5 are shown in Table Substantial interobserver agreement among the participating radiologists was found by means of an intra-class correlation coefficient of 0.67 (p < 0.001) for the VGAS. As a result, image quality scores were averaged over the three readers and shown in Figure

104 For the standard scan (protocol 3), the average score for bone detail and image artifacts increased and the average score for confidence in grading bone loosening and ossification stayed the same when SEMAR was utilized. For low-dose scans (protocols 4 through 6), the average score for bone detail and image artifacts and the average score for confidence in grading bone loosening and ossification stayed the same when SEMAR was utilized. For the high-dose scans (protocols 1 and 2), the average score for bone detail and image artifacts increased and the average score for confidence in grading bone loosening and ossification stayed the same when SEMAR was utilized. When it came to scoring the soft tissue features, only one of the three musculoskeletal radiologists felt comfortable evaluating the soft tissue. The other two radiologists stated the edema in the tissue reduced the soft contrast visualization and they did not feel confident in scoring the features. For the standard scan (protocol 3), the image quality scores for soft tissue contrast, confidence in diagnosing a bladder mass, and confidence in diagnosing a hematoma increased and the image quality scores for soft tissue noise stayed the same when SEMAR was utilized. For low-dose scans (protocols 4 through 6), image quality scores for soft tissue contrast and soft tissue noise stayed the same (33.3%) or increased (66.6%) when SEMAR was utilized, and confidence in diagnosing a bladder mass or hematoma increased when SEMAR was utilized. For the high-dose scans (protocols 1 and 2), image quality scores for soft tissue contrast stayed the same (50%) or improved (50%), image quality scores for soft tissue noise stayed the same, and image quality scores for confidence in diagnosing a bladder mass or hematoma improved when SEMAR was utilized. 104

105 A reduction in image quality was seen for bone detail, confidence in grading bone loosening and ossification, soft tissue contrast, and soft tissue noise when scanned with the lowest dose protocol, protocol 6. However, SEMAR still produced the same (40%) or improved (60%) image quality scores in these cases, as compared to reconstructions without SEMAR. Furthermore, when scanned with a higher dose protocol (protocol 1) improvements in image quality were only observed for soft tissue contrast, with all other features producing the same average scores as protocol Discussion Although 300 patients is a small sample size compared to the number of patients who undergo CT examinations per year, we were limited to the time-consuming task of manually scrolling through each patient s scan series. Future studies should conduct the assessment with a greater sample size. Furthermore, due to the small sample size, the study was limited to adults older the age of 50, who were more likely to have metal implants than younger adults. In the investigation with one cadaver containing metal implants in the sternal, lumbar, and pelvic regions, a dose increase was observed relative to two cadavers without metal implants, with a 56% increase in effective mas and a 33% maximum increase in organ dose. The effective mas was affected by the metal implants and the arms being at the side of the head. Therefore, it is likely the increase in effective mas would have been less than 56% if the extended arms were not in the field of view. A similar study by Rizzo et al compared mean effective mas in 40 patients with and without metal hip implants, and found that for patients weighing between lbs, the effective mas increased by 34.1%, on average (91). 105

106 In our study, the organ doses increased less than 4 mgy, reaching a maximum increase of 3.2 mgy in the small intestine. This is not a significant concern in terms of stochastic risks, especially for older patients above the age of 50. However, the tube current did not reach a maximum value of 500 ma, but may have been higher with a scanner not equipped with iterative reconstruction, resulting in increased organ doses. In order to assess the effect of metal on image quality, it was important to evaluate image quality in regions containing no metal implants and in regions containing large amounts of metal implants. Although the best assessment would be to evaluate the same region with and without metal implants, this would require obtaining postmortem subjects without metal implants, imaging them, having metal implants surgically placed within them, and imaging them again. This approach requires an extensive amount of work, requiring for a medical professional to place a hip prosthesis within an embalmed postmortem subject. Being that we already had subjects with previously inserted prostheses, we determined it was feasible to evaluate image quality in pelvis regions superior to the hip implants and directly at the hip implants. This study only investigated image artifacts and diagnostic confidence in subjects with hip prosthesis. However, we found a high prevalence of patients presenting with metal implants in the spine and sternum. It is of interest to conduct the same type of investigation, using a task-driven image quality evaluation for subjects or patients containing metal implants in other anatomic areas. 106

107 Containing metal implants 27% No metal implants 73% Figure 6-1. Prevalence of patients containing metal implants in 300 adults examined with CT at UF Health Other, 19 Hip replacement, 17 Pacemaker, 6 Cardiac Graft, 6 Fixation hardware, 34 Figure 6-2. Types and number of metal implants encountered in 82 patients at UF Health 107

108 Table 6-1. Scan techniques for standard CAP protocol utilized on Subjects 1, 2, and 3 Scan Techniques Subject 1 Subject 2 Subject 3 Acquisition Helical Helical Helical Tube Voltage (kvp) Slice thickness (mm) Number of detectors Rotation time (s) Pitch Focal spot size Small Small Small Bow-tie filter Large Large Large Target noise level (SD) ma Range Effective mas (mas) CTDIvol (mgy) DLP (mgy-cm) Note: Tube current was modulated with Sure Exposure 3D and dose reduction was achieved with iterative reconstruction algorithm AIDR-3D A B Figure 6-3. Graphs of tube current modulating throughout the scan length of Subjects 1, 2, and 3. The tube currents are superimposed on a scanogram of Subject 3 in the (A) AP and (B) LAT orientation. Subjects 1, 2, and 3, are represented by the green, purple, and orange lines, respectively. Figure courtesy of author. 108

109 Organ Dose (mgy) Table 6-2. Organ doses measured with CAP CT scans of Subject 1, 2, and 3. Organ Dose (mgy) Subject 1 Subject 2 Subject 3 Organs Avg Max SD Avg Max SD Avg Max SD Lung Liver Stomach SI Colon Ovary Note: Organ doses are reported as the average (Avg), maximum (Max) and standard deviation (SD) of all organ dose measurements acquired in each organ. 14 Subject 1 Subject 2 Subject Lung Liver Stomach Small Intestine Colon Ovary Figure 6-4. Average organ doses measured with CAP CT scans of Subjects 1, 2, and 3 109

110 Table 6-3. Scan techniques for standard adult CAP protocol utilized on Subject 3 Scan Techniques Subject 3 Acquisition Helical Tube Voltage (kvp) 120 Slice thickness (mm) 0.5 Number of detectors 80 Rotation time (s) 0.5 Pitch Focal spot size Bow-tie filter Small Large Target noise level (SD) 12.5 ma Range Effective mas (mas) 227 CTDIvol (mgy) 9.7 DLP (mgy-cm) Note: Tube current was modulated with Sure Exposure 3D and dose reduction was achieved with iterative reconstruction algorithm AIDR-3D Table 6-4. Radiologists utilized to assess image quality of a CAP CT scan of Subject 3 Radiologist ID Radiologist Specialty Radiologist Name 1 Pediatric Jonathan Williams 2 Pediatric Dhanashree Rajderkar 3 Pediatric Robert Dubuisson 4 Thoracic Tan-Lucien Mohammed 5 Thoracic Robbie Slater 6 Abdomen Suzanne Mastin 7 Abdomen Eric Thoburn 8 Resident Ashish Sethi 9 Resident Brandon Custer 10 Resident Zachary Schirm 11 Resident Garret Woodbury 110

111 Table 6-5. Scan techniques for standard adult PWH protocol utilized on Subject 3 Scan Techniques Value Acquisition Tube Voltage (kvp) 135 Slice thickness (mm) 0.5 Number of detectors 320 Rotation time (s) 0.5 Focal spot size Volumetric Small Field of View (mm) Bow-tie filter Large Scan Range (mm) 160 Target noise level (SD) 12.5 Effective mas (mas) 250 CTDIvol (mgy) 8.9 DLP (mgy-cm) Note: Tube current was modulated with Sure Exposure 3D and dose reduction was achieved with iterative reconstruction algorithm AIDR-3D Table 6-6. Radiologists utilized to assess image quality of a PWH CT scan of Subject 3 Radiologist ID Radiologist Specialty Radiologist Name 1 Musculoskeletal Cooper Dean 2 Musculoskeletal Charles Bush 3 Musculoskeletal Troy Storey 4 Thoracic Tan-Lucien Mohammed 5 Resident James Freeman 111

112 Table 6-7. Scan techniques for standard adult PWH protocol utilized on Subject 5 Scan Techniques Value Acquisition Tube Voltage (kvp) 135 Slice thickness (mm) 0.5 Number of detectors 320 Rotation time (s) 0.5 Focal spot size Volumetric Small Field of View (mm) Bow-tie filter Large Scan Range (mm) 160 Target noise level (SD) 12.5 Effective mas (mas) 45 CTDIvol (mgy) 4.7 DLP (mgy-cm) 75.7 Note: Tube current was modulated with Sure Exposure 3D and dose reduction was achieved with iterative reconstruction algorithm AIDR-3D Table 6-8. Radiologists utilized to assess image quality of a PWH CT scan of Subject 5 Radiologist ID Radiologist Specialty Radiologist Name 1 Musculoskeletal Cooper Dean 2 Musculoskeletal Charles Bush 3 Musculoskeletal Troy Storey 112

113 VGAS Table 6-9. Image quality scores of Subject 3 scanned with a CAP CT protocol Image Quality Feature Radiologist ID Lung detail in chest Soft tissue visualization in abdomen Soft tissue visualization in pelvis Note: All features were evaluated by radiologists listed in Table Lung Detail in Chest Soft Tissue Visualization in Abdomen Soft Tissue Visualization in Pelvis Figure 6-5. Image quality scores of Subject 3 scanned with a CAP CT protocol. Image quality scores were averaged across radiologists listed in Table 6-4 to calculate the visual grading analysis score (VGAS). Protocol techniques are described in Table

114 VGAS Table Image quality scores of Subject 3 scanned with a PWH CT protocol in regions of the pelvis with and without metal implants Region in the Pelvis Without Metal Implants With Metal Implants Image Quality Feature Radiologist ID Radiologist ID Soft Tissue Visualization Bone Detail Note: The scan techniques utilized for the PWH CT protocol utilized on Subject 3 are listed in Table 6-5. All features were evaluated by radiologists listed in Table 6-6. Without Metal Implants With Metal Implants Soft Tissue Visualization Bone Detail Figure 6-6. Image quality scores of Subject 3 scanned with a PWH CT scan by image quality feature and area imaged. Image quality scores were averaged across radiologists listed in Table 6-6 to calculate the visual grading analysis score (VGAS). Protocol techniques are listed in Table

115 VGAS Table Image quality scores of Subject 5 scanned with a PWH CT protocol in regions of the pelvis with and without metal implants Region in the Pelvis Without Metal Implants With Metal Implants Image Quality Feature Radiologist ID Radiologist ID Bone detail Confidence in grading loosening Confidence in grading ossification Image artifacts Soft tissue contrast Soft tissue noise Note: The scan techniques utilized for the PWH CT protocol utilized on Subject 5 are listed in Table 6-7. All features were evaluated by radiologists listed in Table Without Metal Implants With Metal Implants Bone detail Confidence in grading loosening Confidence in grading ossification Image artifacts Soft tissue contrast Soft tissue noise Figure 6-7. Image quality scores of Subject 5 scanned with a PWH CT scan by image quality feature and area imaged. Image quality scores were averaged across radiologists listed in Table 6-8 to calculate the visual grading analysis score (VGAS). Protocol techniques are listed in Table

116 A B C D Figure 6-8. Pelvic images of Subject 3 reconstructed with and without SEMAR. Images were reconstructed with a 1-mm smooth soft tissue algorithm, without (A) and with (B) the SEMAR algorithm and with a 0.5 mm sharp bone algorithm, without (C) and with (D) the SEMAR algorithm. Figure courtesy of author. A B C D Figure 6-9. Pelvic images of Subject 5 reconstructed with and without SEMAR. Images were reconstructed with a 2-mm smooth soft tissue algorithm, without (A) and with (B) the SEMAR algorithm and with a 0.5 mm smooth soft tissue algorithm, without (C) and with (D) the SEMAR algorithm. Figure courtesy of author. 116

117 Image Quality Score Image Quality Score Table Image quality scores of Subject 3 scanned with a PWH CT protocol with and without SEMAR Metal Artifact Reduction None SEMAR Image Quality Feature Radiologist ID Radiologist ID Soft Tissue Visualization Bone Detail Note: Protocol techniques are listed in Table 6-5. All features were assessed in the pelvis region containing metal implants by radiologists listed in Table 6-6 None SEMAR 3 Soft tissue visualization 3 Bone detail A Radiologist ID Radiologist ID B Figure Image quality scores for pelvis scans of Subject 3 by image quality feature, radiologist ID, and metal artifact reduction (SEMAR). Protocol techniques are listed in Table 6-5. Features were assessed in the pelvis region containing metal implants by radiologists listed in Table

118 VGAS Table Image quality scores of Subject 5 scanned with a PWH CT protocol without and with SEMAR Image Quality Feature Metal Artifact Reduction None Radiologist ID SEMAR Radiologist ID Bone detail Confidence in grading loosening Confidence in grading ossification Image artifacts Soft tissue contrast Soft tissue noise Confidence in diagnosing a bladder mass Confidence in diagnosing a hematoma Note: All features were assessed in the pelvis region containing metal implants by radiologists listed in Table None SEMAR Bone detail Confidence in grading loosening Confidence in grading ossification Image artifacts Soft tissue contrast Soft tissue noise Confidence Confidence in diagnosingin diagnosing a bladder a hematoma mass Figure Image quality scores of Subject 5 scanned with a PWH CT scan by image quality feature and metal artifact reduction (SEMAR). Image quality scores were averaged across radiologists listed in Table 6-8 to calculate the visual grading analysis score (VGAS). Protocol techniques are listed in Table

119 Table Scan techniques for modified PWH protocols acquired with metal artifact reduction on Subject 3 Scan Protocol Tube Voltage (kvp) Target Noise Level (SD) Effective mas CTDIvol (mgy) Table Scan techniques for modified PWH protocols acquired with metal artifact reduction on Subject 5 Scan Protocol Tube Voltage (kvp) Target Noise Level (SD) Effective mas (mas) CTDIvol (mgy)

120 Table Image quality scores of Subject 3 scanned with modified PWH CT protocols with and without SEMAR Metal Artifact Reduction None SEMAR Image Quality Feature Soft Tissue Visualization Scan Protocol Radiologist ID Radiologist ID Bone Detail Note: Scan protocols are described in Table Features were assessed in the pelvis region containing metal implants by radiologists listed in Table

121 Image Quality Score Image Quality Score Image Quality Score Image Quality Score Image Quality Score None SEMAR Scan Protocol Scan Protocol A B C Scan Protocol Scan Protocol D E Scan Protocol Figure Image quality scores for soft tissue visualization of pelvis scans with and without SEMAR in Subject 3 for A) Radiologist 1, B) Radiologist 2, C) Radiologist 3, D) Radiologist 4, and E) Radiologist 5. Scan protocols are described in Table Features were assessed in the pelvis region containing metal implants by radiologists listed in Table

122 Image Quality Score Image Quality Score Image Quality Score Image Quality Score Image Quality Score None SEMAR A Scan Protocol Scan Protocol B C Scan Protocol Scan Protocol D E Scan Protocol Figure Image quality scores for bone detail of pelvis scans with and without SEMAR in Subject 3 for A) Radiologist 1, B) Radiologist 2, C) Radiologist 3, D) Radiologist 4, and E) Radiologist 5. Scan protocols are described in Table Features were assessed in the pelvis region containing metal implants by radiologists listed in Table

123 Table Image quality scores of Subject 5 scanned with modified PWH CT protocols without and with SEMAR Metal Artifact Reduction None SEMAR Image Quality Feature Bone Detail Scan Protocol Radiologist ID Radiologist ID Confidence in grading bone loosening Confidence in grading bone ossification Image artifacts

124 Table Continued Metal Artifact Reduction None SEMAR Image Quality Feature Soft Tissue Contrast Scan Protocol Radiologist ID Radiologist ID Soft tissue noise Confidence in diagnosing a bladder mass Confidence in diagnosing a hematoma Note: Scan protocols are described in Table Features were assessed in the pelvis region containing metal implants by radiologists listed in Table

125 VGAS VGAS VGAS VGAS VGAS VGAS VGAS VGAS 3 Bone Detail None SEMAR 3 Image artifacts A Scan Protocol Scan Protocol B Confidence in grading bone Confidence in grading bone 3 loosening 3 ossification C Scan Protocol Scan Protocol D Soft tissue contrast Soft tissue noise Scan Protocol Scan Protocol E F Confidence in diagnosing a Confidence in diagnosing a bladder mass hematoma G Scan Protocol Scan Protocol H Figure Image quality scores for PWH scans of Subject 5 by image quality feature, scan protocol, and metal artifact reduction (SEMAR). Image quality scores were averaged across radiologists listed in Table 6-8 to calculate the visual grading analysis score (VGAS). Protocol techniques are listed in Table

126 CHAPTER 7 PROTOCOL OPTIMIZATION IN LUNG CANCER SCREENING CT 7.1 Lung Cancer Screening in CT In the United States, lung cancer is the leading cause of death from cancer (92). The National Cancer Institute estimated that in 2013, there were over 220,000 new cases of lung cancer, and over 150,000 deaths from lung cancer, exceeding the total number of deaths from colon, breast, and prostate cancer combined (93). These numbers can be explained by the fact that the majority of lung cancers are diagnosed after the cancer has spread to distant regions, with a five-year survival rate of only 4 percent (93). Alternatively, detecting lung cancer at its earliest stage increases the fiveyear survival rate closer to 57 percent, driving much research to focus on identifying effective methods of early detection. Technological developments in computed tomography have allowed for highresolution volumetric images to be obtained in a single breath hold, while utilizing lower levels of radiation, extending its use into lung applications (94). In 2002, the National Cancer Institute funded the National Lung Screening Trial (NLST) to determine whether lung cancer mortality in a high-risk population could be reduced by screening with low-dose CT (LDCT) as compared with a single-view chest radiograph (CXR) (95). From 2002 to 2007, 53,454 participants, ranging from 55 to 74 years of age with a significant smoking history of at least 30 pack-years, were randomly assigned to undergo three annual lung cancer screening exams with either standard CXR or helical chest LDCT. During the trial, 26,724 participants received lung cancer screening exams with LDCT administered on 97 scanners at 33 sites nationwide. 126

127 In June 2011, the NLST reported that 20% fewer lung cancer deaths were seen in participants screened with LDCT compared to those who were screened with CXR (96). Specifically, after accruing data from 287,471 person-years, there were 356 deaths from lung cancer in the LDCT group and 443 deaths from lung cancer in the CXR group. Since then, the U.S. Preventive Services Task Force (USPSTF) recommended lung cancer screening with LDCT, and the Centers for Medicare and Medicaid Services (CMS) added lung cancer screening with LDCT as an additional preventative service benefit for eligible patients (97). The AAPM Working Group on Standardization of CT Nomenclature and Protocol developed recommended protocols based on manufacturers LDCT Chest protocols as well as the group s experience with the NLST (98). The first part of this chapter measures typical organ doses resulting from for LDCT lung cancer screening exams; the second part aims to identify an optimal reconstruction setting for LDCT lung cancer screening exams based on radiologist preference; the third part investigates the potential for further dose reduction in LDCT lung cancer screening exams; the fourth part measures organ doses resulting from standard and reduced-dose LDCT lung cancer screening exams; and the final part investigates patient dose and clinical findings in patients examined with LDCT lung cancer screening exams at UF Health. 7.2 Measuring Organ Doses from Lung Cancer Screening CT Despite decreasing trends in smoking, the population at risk for lung cancer continues to be large, with an estimated 43.4 million current cigarette smokers and 47.3 million former cigarette smokers in the United States (99). About 8 million of these are in the target age group of years old and have smoked at least a pack a day for

128 years (76). With a large population recommended to undergo annual lung cancer screening, an organ dose assessment was required to consider the risk-benefit of LDCT lung cancer screening Methods Subjects 5 and 7 were scanned with an AAPM-recommended protocol for a Toshiba Aquilion ONE 320-slice CT scanner, with helical acquisition, from the top to the bottom of the lungs, 120 kvp, 0.5 x 80 detectors, 0.35 s rotation time, pitch, tube current modulating between 10 and 150 ma with a target noise level of 20 SD (98). Organ doses were measured in Subject 5 with 2 OSL dosimeters in the thyroid, 8 in the lungs, 8 in the breasts, 3 in the liver, 2 in the stomach, 1 in the small intestine, 2 in the ovaries, 1 in the uterus, and 15 across the skin. Subject 7 only had 1 dosimeter in the thyroid, with all other organs having the same number of dosimeters as Subject Results Scanning with the AAPM-recommended protocol resulted in a CTDIvol of 2.2 mgy for Subject 5 and 5.3 mgy for Subject 7. This trend in CTDIvol was expected, as Subject 5 was smaller in size than Subject 7, where Subject 5 had a BMI of kg/m 2 and an effective diameter of cm in the chest, and Subject 7 had a BMI of kg/m 2 and an effective diameter of cm in the chest. The average, maximum, and standard deviation values of measured organ doses are listed in Table 7-1 for Subject 5 and Table 7-2 for Subject 7. For the superior organs located in the scan field of view, Subject 5 received reduced organ doses than Subject 7. Compared to the superior organs, the inferior organs outside of the field of view, such as the small intestine, colon, ovaries, and uterus received lower organ doses below 2 128

129 mgy due to internal scattered radiation from the superior organs. The average organ doses for Subjects 5 and 7 are graphed in Figure Optimizing Image Reconstruction in Lung Cancer Screening CT The clinical Chest CT protocol implemented at UF Health is utilized for a variety of indications, including evaluation of chest abnormalities, trauma, lung cancer, and more. The protocol produces a lung image series reconstructed with 2-mm axial slices using a Lung Sharp algorithm, as well as a soft tissue image series reconstructed with 3-mm axial slices using a Standard Body Axial algorithm. These reconstruction settings were initially selected for the LDCT lung cancer screening protocol at UF Health. However, AAPM recommends reconstructing with thinner axial slices of 1-mm for both the lung and soft tissue reconstructions (98). A thinner slice acquisition yields better image detail, but also results in greater image noise. Nowadays, scanners acquire thin slices of 0.5 mm but reconstruct thicker images of 1-3 mm for interpretation. The thicker images have a better SNR, while thin images are still used to look at critical details and to get 2D reformation and 3D analysis. In addition to the Lung Sharp algorithm, the Toshiba Aquilion ONE CT scanner is equipped with two additional lung reconstruction algorithms called Lung High Resolution CT and Lung Low Dose. The Lung High Resolution CT algorithm is used to evaluate pulmonary diseases such as bronchiectasis, and the Lung Low Dose algorithm is indicated for follow up of known disease. With the reconstruction algorithms being proprietary, it was not possible to predict how each algorithm affects image appearance, requiring clinical investigation using real anatomy. With a variety of reconstruction combinations of slice thicknesses and algorithms available, it was 129

130 necessary to identify the optimal reconstruction parameters for lung cancer screening CT Methods Subject 4 was scanned with a standard LDCT lung cancer screening protocol, and images were reconstructed with one soft tissue algorithm (Standard Body Axial, FC18) and three lung algorithms (Lung Sharp Chest, FC86; Lung Low Dose, FC55; and Lung High Resolution CT, FC86) with two different slice thicknesses (1-mm and 3-mm). In total, 2 soft tissue and 6 lung image series were reconstructed with the same raw data of a standard lung cancer screening CT. The Chief of Thoracic Imaging, Dr. Mohammed, was blinded to reconstruction parameters and was guided through a series of questions to identify the preferred reconstruction setting for lung cancer screening CT Results Figure 7-2 displays axial chest images reconstructed with the Standard Body Axial algorithm at 1-mm and 3-mm slice thicknesses. Upon viewing the entire soft tissue image series at the two different slice thicknesses of 1-mm and 3-mm, Dr. Mohammed stated that the 1-mm image appeared too noisy and he preferred the 3-mm reconstruction due to better SNR. Figure 7-3 displays axial lung images reconstructed with the Lung Sharp algorithm at 1-mm and 3-mm slice thicknesses. Upon viewing the entire lung image series at the two different slice thicknesses of 1-mm and 3-mm, Dr. Mohammed stated that the detail of the lung parenchyma looked sharper with 1-mm slices, despite the increase in noise. For the purposes of identifying lung nodules, he preferred the 1-mm slice thickness for the lung reconstruction. 130

131 Figure 7-4 displays axial lung images reconstructed with the Lung Sharp, Lung Low Dose, and Lung High Resolution CT algorithms at 1-mm slice thicknesses. When comparing the lung image reconstruction algorithms, Dr. Mohammed stated that although the Lung Low Dose looked similar to Lung Sharp, he preferred the appearance of lung detail with the Lung Sharp algorithm. He also found that the Lung High Resolution CT algorithm looked much noisier than Lung Sharp, and the vessels appeared smoother with the Lung Sharp algorithm. It was decided that the ideal reconstruction parameter would be the Standard Body Axial algorithm at 3 mm and the Lung Sharp algorithm at 1 mm. 7.4 Assessing Image Quality for Reduced-Dose Lung Cancer Screening CT While radiation dose should be minimized to reduce long-term cancer risks, it also directly affects CT image quality and the ability to visually diagnose small nodules. With the nationwide acceptance of the LDCT lung cancer screening initiative, it remained necessary to identify an adequate protocol that considers both radiation dose and image quality. The purpose of this study was to assess image quality for reduceddose LDCT lung cancer screening exams Methods Subjects 3, 4, 5, 7, 8, 9, 10, and 11 were scanned with several LDCT lung cancer screening protocols. Because the chest CT protocol utilized a tube voltage of 120 kvp, initially we were interested in achieving dose reduction by only reducing the tube current. Therefore, Subject 3 was scanned with a tube voltage of 120 kvp, with the target noise level modified from 20 SD to 25 SD. The scan parameters utilized for Subject 3 are listed in Table 7-3. By the time the next postmortem subject was utilized, we were interested in investigating the dose reduction and image quality achieved with 131

132 a lower tube voltage. Therefore, Subject 4 was scanned with tube voltages of 120 kvp and 100 kvp, and target noise levels of 20 SD and 25 SD. The scan parameters utilized for Subject 4 are listed in Table 7-4. Once the image quality analysis was conducted, and we learned that radiologists found some of the protocols with 25 SD to be acceptable, we increased our investigation with a higher target noise level to verify whether lower tube current would be acceptable. Subject 5 was scanned with a higher target noise level of 35 SD, with scan parameters listed in Table 7-5. Once we received Subject 7, the tube rotation time was updated from 0.5 s to 0.35 s on the AAPM recommendations for the Toshiba Aquilion ONE Scanner. As a result, Subject 7 was scanned with the newly updated rotation time of 0.35 s. Furthermore, once we learned that images of Subject 5 acquired with 35 SD were too noisy, we reduced the target noise level down to 30 SD for Subject 7, with scan parameters listed in Table 7-6. After Subject 7 had been utilized, our team had the opportunity to borrow 4 cadavers from the medical school anatomy lab to scan for image quality analysis. As a result, four more subjects, Subjects 8, 9, 10, and 11, were scanned with several LDCT lung cancer screening protocols with scan parameters listed in Table 7-7, Table 7-8, Table 7-9, and Table 7-10, respectively. Although the first part of this study identified 1-mm slices to provide the best lung detail, CT management decided to reconstruct clinical lung images to 2-mm until the PACS could handle the increase in data. As a result, the image quality analysis was carried out using 2-mm axial slices with a Lung Sharp algorithm in order to closely match clinical implementation. Two board certified thoracic radiologists with experience in reading lung cancer screening exams, listed in Table 7-11, graded the sharp reproduction of lung detail, 132

133 diagnostic confidence in assessing lung cancer, and diagnostic confidence in assessing lung disease using a three-point scale of 3 for acceptable image quality, 2 for borderline acceptable image quality, and 1 for unacceptable image quality. For Subjects 8, 9, 10, and 11, an additional question of confidence in visualizing an existing nodule was added to the image quality analysis. The radiologists stated that different lung diseases were present in each of the subjects, requiring fine detail to diagnose the disease. Because small nodules were not always present, the detail required for lung disease was used as a surrogate for fine detail. All subjects presented with some sort of lung disease, either from when the subject was living or from embalming effects, with the exception of Subject 11, who did not present with lung disease. Therefore, Subject 11 did not receive scores for confidence in assessing lung disease Results Subject 3 presented with signs of cystic lung disease. Image quality scores for Subject 3 are listed in Table The default lung cancer screening protocol (protocol 1) received all acceptable image quality scores, with the exception of the sharp reproduction of lung detail scored as borderline acceptable to Dr. Verma. She stated that the collapsed lung made it difficult to visualize the fine detail, but this was likely due to the subject being deceased, and not the scan protocol. The reduced-dose protocol with 120 kvp and 25 SD (protocol 2) received an unacceptable image quality score for sharp reproduction of lung detail, and a borderline acceptable image quality score for confidence in diagnosing nodules by Dr. Verma. Dr. Mohammed graded all features to be acceptable for both scan protocols. Subject 4 presented with a 5-mm granuloma in the lower right lobe and intralobular septal thickening. The image quality scores for Subject 4 are listed in Table 133

134 7-13. The default lung cancer screening protocol (protocol 1) received all acceptable image quality scores by both radiologists. Both of the reduced-dose protocols acquired with 100 kvp (protocols 3 and 4) received borderline acceptable image quality scores for the sharp reproduction of lung detail by both radiologists. The protocol acquired with 120 kvp and 25 SD (protocol 2) produced blurry lung detail, where both radiologists scored the sharp reproduction of lung detail to be of borderline acceptable image quality. All protocols received acceptable image quality scores for diagnostic confidence in diagnosing nodules, with the exception of the lowest dose protocol (protocol 4) which received a borderline acceptable score by Dr. Mohammed. Subject 5 presented with a 9-mm granuloma in the upper right lobe, and signs of bronchiectasis in the lower right lobe. Figure 7-5 displays CT images of Subject 5 acquired with the highest and lowest dose protocols in the axial slices that contain the 9-mm granuloma. In the lowest dose protocol, the granuloma presented with blurred margins and the lung detail was scored as having suboptimal image quality. The image quality scores for Subject 5 are listed in Table The scans acquired with a target noise level of 25 SD (protocols 2 and 5) did not result in any unacceptable or borderline acceptable scores, and had identical image quality scores to that of the default protocol (protocol 1). The scans acquired with a target noise level of 35 SD (protocols 3 and 6) resulted in borderline acceptable scores, and were deemed to be unsuitable for lung cancer screening. Image quality scores for Subject 7 are listed in Table Protocols with 100 kvp and 120 kvp acquired with a target noise level of 20 and 25 SD (protocols 1, 2, 5, and 6) received acceptable image quality scores for all image quality features by both 134

135 radiologists. The protocols acquired with a target noise level of 30 SD and 35 SD (protocols 3, 4, 7, and 8) received borderline acceptable scores by both radiologists for the sharp reproduction of lung detail. Furthermore, the lung disease was compromised for the two protocols acquired with 35 SD (protocols 4 and 8), receiving borderline acceptable scores by both radiologists for confidence in diagnosing lung disease. Subject 8 presented with a 3-mm granuloma in the right lobe. Figure 7-6 displays CT images of Subject 8 acquired with the highest and lowest dose protocols in the axial slices that contain the 3-mm granuloma. For both the highest and lowest dose protocols, the visualization of the 3-mm nodule received acceptable scores by both radiologists. However, they stated that it is due to its high contrast appearance, being a calcified nodule, and they were not confident a low contrast nodule would be as clearly depicted. The image quality scores for Subject 8 are listed in Table Only the protocols with 35 SD (protocols 4 and 8) received borderline acceptable image quality scores. The scan protocol with 120 kvp with 35 SD (protocol 4) received borderline acceptable scores by both radiologists for sharp reproduction of lung detail. The scan protocols with 35 SD (protocol 4 and protocol 8) received borderline acceptable scores by Dr. Mohammed for confidence in diagnosing nodules and confidence in diagnosing lung disease. Subject 9 presented with two large speculated nodules with 19-mm and 14-mm diameters. Dr. Mohammed stated that it is likely the patient died of lung cancer. Figure 7-7 displays the larger 19-mm speculated nodule in the right lobe with the highest and lowest dose protocols (protocols 1 and 8). The visualization of the existing nodule was acceptable in both protocols, however the lowest dose protocol received borderline 135

136 acceptable scores for the sharp reproduction of lung detail. Image quality scores for Subject 9 are listed in Table The scan acquired with 120 kvp with 35 SD (protocol 4) received borderline acceptable image quality scores by Dr. Verma for sharp reproduction of lung detail and confidence in diagnosing nodules. The scan acquired with 100 kvp with 30 SD (protocol 7) received a borderline acceptable score by Dr. Mohammed for sharp reproduction of lung detail. The lowest dose scan acquired with 100 kvp with 35 SD (protocol 8) received borderline acceptable scores by both radiologists for sharp reproduction of lung detail, and received a borderline acceptable score by Dr. Mohammed for confidence in diagnosing nodules. Subject 10 had a 6-mm pulmonary nodule and dilated air space. Image quality scores for Subject 10 are listed in Table Protocols acquired with 120 kvp with 20, 25, and 30 SD (protocols 1, 2, and 3) and protocols acquired with 100 kvp with 20 and 25 SD (protocols 5 and 6) received all acceptable image quality scores. Protocols acquired with both 120 kvp and 100 kvp with 35 SD (protocols 4 and 8) received borderline acceptable scores by both radiologists for sharp reproduction of lung detail and a borderline acceptable score by Dr. Verma for confidence in diagnosing nodules. Dr. Mohammed scored the 100 kvp 35 SD protocol (protocol 8) to have borderline acceptable visualization of an existing nodule. The protocol acquired with 100 kvp and 35 SD (protocol 7) received borderline acceptable scores by both radiologists for sharp reproduction of lung detail. Image quality scores for Subject 11 are listed in Table All protocols acquired with 120 kvp and 100 kvp with 20 and 25 SD (protocols 1, 2, 5, and 6) and with 100 kvp with 30 SD (protocol 7) received acceptable image quality. The protocol 136

137 acquired with 120 kvp with 30 SD (protocol 3) received borderline acceptable scores by Dr. Verma for the sharp reproduction of lung detail and confidence in diagnosing nodules. The protocols acquired with 120 kvp and 100 kvp with 35 SD (protocols 4 and 8) received borderline acceptable scores by both radiologists for sharp reproduction of lung detail. Dr. Verma scored the 120 kvp protocol with unacceptable image quality and the 100 kvp protocol with borderline acceptable image quality for the confidence in diagnosing nodules. A protocol- and radiologist-specific VGAS was calculated using image quality scores for Subjects 3, 4, 5, 7, 8, 9, 10, and 11. Only Subjects 7 through 11 were scanned with 8 similar scan protocols. Subjects 3, 4, 5, and 7 were scanned with fewer protocols, therefore had their scores averaged into the protocols techniques that they were examined with. Figure 7-8 shows the VGAS for the eight different scan protocol. Overall, all scan protocols that were acquired with a target noise level of 35 SD for six subjects received borderline acceptable scores for at least one of the image quality features. This was attributed to the tube current reduction leading to an inappropriate level of image noise and affecting the appearance of the lung. For five subjects that were scanned with a target noise level of 30 SD, one subject (Subject 7) received borderline acceptable scores for both scan protocols acquired with 100 kvp and 120 kvp. Two subjects (Subject 8 and Subject 11) received acceptable image quality scores for scan protocols acquired with 100 kvp, and three subjects (Subject 8, Subject 9, and Subject 10) received acceptable image quality scores for scan protocols acquired with 120 kvp. 137

138 For eight subjects that were scanned with a target noise level of 25 SD, six subjects (Subject 5, Subject 7, Subject 8, Subject 9, Subject 10, and Subject 11) received acceptable image quality scores for both 120 kvp and 100 kvp. Two subjects (Subject 3 and Subject 4) received borderline acceptable scores for the sharp reproduction of lung detail and confidence in diagnosing nodules. However, subjects 3 and 4 had collapsed lungs, which likely attributed to the poor image quality and visualization of the lung detail. It was determined that the scan protocols with 25 SD produced sufficient acceptable scores, and was a suitable protocol to propose for use in patients. 7.5 Measuring Organ Doses from Reduced-Dose Lung Cancer Screening CT Methods With acceptable scan protocols identified by the radiologists in Section 6.3.2, organ doses were measured in Subjects 5 and 7 for the acceptable reduced-dose lung cancer screening protocols with 120 kvp and 20 SD, 120 kvp and 25 SD, 100 kvp and 20 SD, and 100 kvp and 25 SD. The scan techniques utilized for the four acceptable protocols are listed in Table 7-20 for Subject 5 and Table 7-21 for Subject Results The average, maximum, and standard deviation values of measured organ doses for the standard and reduced-dose protocols are listed in Table 7-22 for Subject 5 and Table 7-23 for Subject 7. Average organ doses are graphed in Figure 7-9 for Subject 5 and Figure 7-10 for Subject 7. Percent organ dose reductions were calculated for each of the reduced-dose protocols relative to the reference scan (protocol 1) and listed in Table 7-24 for Subject 5 and Table 7-25 for Subject

139 For Subject 15, increasing the target noise level from 20 to 25 SD yielded the greatest dose savings to the organs directly in the scan range, with individual organ dose reductions ranging from 23% to 37%. The average organ dose savings for organs that were within the scan range was 30%, 14%, and 24% for protocols 2, 4, and 6, respectively. For Subject 17, reducing the tube current from 120 to 100 kvp and increasing the target noise level from 20 to 25 SD yielded the greatest dose savings to the organs directly in the scan range, with individual organ dose reductions ranging from 32% to 38%. The average organ dose savings for organs that were within the scan range was 20%, 34%, and 35% for protocols 2, 5, and 6, respectively. The tube current modulates based on the target noise level specified by the user and the attenuation of the patient in the x,y, and z-axis, driven by the automated tube current modulation software. Figure 7-11 and Figure 7-12 show the tube current modulation plots for the four acceptable scan protocols acquired on Subject 5 and Subject 7, respectively. In Figure 7-11, for the protocol with 120 kvp that required the lowest image noise with a target noise level of 20 SD (protocol 1), the tube current, shown in yellow, barely reached 100 ma, with reduced tube current in the chest region with air. When the target noise level was increased to 25 SD (protocol 2), permitting more image noise, the tube current dropped below 50 ma and modulated throughout the patient s anatomy, as shown in orange. For the protocol with a lower tube voltage of 100 kvp and lower image noise with a target noise level of 20 SD (protocol 4), the tube current was highest, passing 100 ma, but not reaching the limit of 150 ma, shown in green. When the target noise level was increased to 25 SD (protocol 5), the tube current reduced back down below 100 ma, shown in blue. For 20 SD, the protocol with 100 kvp 139

140 used a higher tube current, shown in green. For 25 SD, the protocol with 100 kvp used a higher tube current, shown in blue. This was also seen in the scan techniques listed on the scanner, where the effective mas was highest for the protocol with 100 kvp and lower target noise level of 20 SD (protocol 4) at 47 mas, and lowest for the protocol with 120 kvp and higher target noise level of 25 SD (protocol 2) at 19 mas. The CTDIvol was highest for the protocol with 120 kvp and 20 SD (protocol 1) at 2.1 mgy, and lowest for the protocol with 120 kvp and 25 SD (protocol 2) at 1.3 mgy. When comparing two protocols with the same target noise level of 20, the protocol with 120 kvp (protocol 1) produced the higher CTDIvol. However, when comparing the two protocols with the same target noise level of 25, the protocol with the lower tube voltage (protocol 5) received the higher CTDIvol of 1.4 mgy, which was still lower than either of the protocols with 20 SD. In Figure 7-12, for the protocol with 120 kvp that required less image noise with a target noise level of 20 SD (protocol 1), the tube current, shown in yellow, nearly maxed out at the limiting tube current of 150 ma, with reduced tube current in the neck. When the target noise level was increased to 25 SD, permitting more image noise, the tube current reduced and modulated throughout the patient s anatomy, as shown in orange. For the protocol with 100 kvp and 20 SD, the tube current maxed out at 150 ma, even through the neck, shown in green. When a higher target noise level was utilized for protocol 6, the tube current dropped in the neck, shown in blue. It can be seen that both protocols that utilized 100 kvp used a higher tube current than both protocols that utilized 120 kvp. This was not as obvious based on the scan techniques listed in Table 7-4, where protocols 1, 5, and 6 received the same effective mas of 64. However, the 140

141 CTDIvol was highest for protocol 1 at 5.3 mgy, and lowest for protocol 6 at 3.2 mgy. When comparing two protocols with the same target noise level, protocols 1 and 5 and protocols 2 and 6, the protocol with 120 kvp always produced a higher CTDIvol. This might be attributed to the larger patient demanding more tube current, with most of the protocols reaching the maximum tube current limit, and therefore the tube voltage having a stronger weight on the dose. However, for Subject 5 who was smaller in size, the tube current never maxed out and the differences in tube current were greater when a lower kvp was utilized, with the tube current having a strong effect on the dose. These differences in ma and CTDIvol values between Subject 5 and Subject 7 highlight that images acquired with higher tube voltage do not necessarily result in higher dose and improved image quality, and the combination of kvp and mas plays a big role into dose and image quality. 7.6 Analysis of Patients Examined with Lung Cancer Screening CT at UF Health The lung cancer screening program was launched at UF Health in March With patient and dose requirements specified by the CMS (97), it was vital to analyze the lung cancer screening examinations performed at UF Health since inception of the program. Specifically, CMS provides reimbursement for annual lung cancer screening for asymptomatic individuals aged 55 to 77 years old, with a smoking history of at least 30 pack-years (where 1 pack-year equals smoking one pack per day for one year), and who are currently smoking or who had quit less than 15 years ago. Furthermore, the CT imaging should preform LDCT with CTDIvol values less than 3 mgy for a standard sized patient, with appropriate reductions and increases in CTDIvol for smaller and larger patients, respectively. Furthermore, with screening performed across multiple CT 141

142 scanners, it was necessary to assess differences in scan parameters and dose across different scanners to ensure techniques were adequately utilized. Clear communication of screening exam results is crucial to guide providers toward appropriate nodule management and to minimize unnecessary workup (100). To this end, the American College of Radiology s (ACR) Lung Imaging Reporting and Data System (Lung-RADS) was created to inform the patient and physicians of findings from lung cancer screening examinations. Inspired by the well-established Breast Imaging Reporting and Data System (BIRADS), Lung-RADS is a standardized reporting system that defines what constitutes a positive screening test, with nodule size and morphology as discriminators (101). A Lung-RADS Category 0 indicates that information is incomplete and additional lung cancer screening CT images or comparison to prior chest CT examinations is needed. Category 1 is reserved for studies that show no nodules or definitely benign nodules. Category 2 nodules have a benign appearance or behavior. The probability of malignancy for category 2 nodules is less than 1%, but does not exclude lung cancer. Examinations in category 1 or category 2 trigger annual followup LDCT in 12 months. Category 3 indicates that a nodule is probably benign, with a low probability that it will become active cancer. An assignment to category 3 triggers follow-up LDCT in 6 months. Category 4 indicates that a nodule is suspicious for malignancy. Management includes additional diagnostic testing with low-dose CT in 3 months, CT with or without contrast, PET/CT, and/or tissue sampling Methods Under institutional review board approval (IRB Study ), we retrospectively accessed the electronic medical record (EMR) and the picture archiving and communication system (PACS) at UF Health to collect and analyze exam 142

143 information for all patients who received lung cancer screening with LDCT at UF Health from March 1, 2015 to October 23, The decision to image was made prior to the retrospective review and was based on current screening guidelines and referral from clinicians. As would be the case at multidisciplinary hospitals that provide several imaging sites, different CT scanners were utilized for the examinations. Scanner profiles included Siemens Sensation 16, Toshiba Aquilion 16, Toshiba Aquilion ONE, and Toshiba Aquilion PRIME. In order to evaluate differences in scan protocols across different scanner models and ensure compliance with CMS requirements, exam details such as date of exam, accession number, location, CT scanner model, CTDIvol and DLP were recorded. Patient demographics such as age, gender, weight, and pack years of smoking were recorded. Furthermore, effective diameter at the central slice in the scan range was measured and recorded to calculate size-specific dose estimate (SSDE) and correlate patient dose with patient size (51). Pertinent findings at screening, such as number and types of nodules and/or masses, and standardized risk assessment category in the form of Lung-RADS, were recorded by a 3 rd year radiology resident, Altan Ahmed, to evaluate the results of the lung cancer screening program at the UF Health Results LDCT lung cancer screening was completed on 109 individuals during the study period. Of these, approximately 64 were female. The median age was 64, with a range years old. The median packs years of smoking were 44 pack years, with a range of pack years. Approximately 103 of individuals met USPSTF guidelines for LDCT screening. Six individuals did not meet the minimum pack years of smoking, ranging 143

144 from 7 to 29 years. The median weight was 173 lbs, with a range of lbs, as shown in Figure The median effective diameter was 29.5 cm, with a range of cm. Figure 7-14 displays the linear correlation of patient weight with BMI and effective diameter in the chest. Eighty-five exams were conducted on two Siemens Sensation 16 CT scanners with 100 kvp, fixed tube current with a ref mas of 50, and a mean CTDIvol of 2.34 mgy with a range of mgy. Sixteen exams were conducted on the Toshiba Aquilion PRIME CT scanner with 120 kvp, modulating tube current with a target noise level of 19 SD, iterative reconstruction, and a mean CTDIvol of 3.2 mgy with a range of mgy. Four exams were conducted on the Toshiba Aquilion 16 slice CT scanner with 100 kvp, modulating tube current with a target noise level of 10 or 15 SD, and a mean CTDIvol of 17.1 mgy with a range of mgy. Two exams were conducted on the Toshiba Aquilion ONE CT scanner with 100 kvp, modulating tube current with a target noise level of 15, iterative reconstruction, and a mean CTDIvol of 9 mgy with a range of mgy. One exam was conducted on the Toshiba Aquilion 64 CT scanner with 100 kvp, tube current modulating with a target noise level of 15, and a CTDIvol of 12.2 mgy. The median CTDIvol was 2.25 mgy, with an average of 3.2 mgy, and a range of mgy. The median SSDE was 2.83 mgy, with an average of 3.8 mgy, and a range of 1.7 to 35.0 mgy. In order to analyze patient dose as a function of patient size, Figure 7-15 shows the CTDIvol and SSDE data for 86 patients who were examined on the two Siemens Sensation 16 CT scanners, as a function of patient effective diameter. For the 86 exams, 81 exams used a fixed tube current setting, delivering an exam CTDIvol of either 144

145 2.25 or 2.5 mgy, independent of patient size, and 5 of the exams used a setting that delivered higher exam CTDIvol values ranging from 2.7 to 4.5 mgy. For the 81 patients that received a fixed ma setting, although the exam CTDIvol was constant independent of patient size, the SSDE had a decreasing trend with patient size, being larger for smaller sized patients. It wasn t until an effective diameter of 35 cm that the SSDE became less than the CTDIvol. The purpose of SSDE is to correct exam CTDIvol to patient size, where a small patient would absorb more dose than a large patient who was exposed to the same amount of radiation. Figure 7-16 shows the CTDIvol and SSDE data for 23 patients who were examined on four Toshiba CT scanners, as a function of patient effective diameter. For the 23 exams, only 16 of them, all conducted on the Toshiba Aquilion PRIME, had an exam CTDIvol below 6 mgy. The remaining 7 patients who were examined on the Toshiba Aquilion ONE, 64 slice, and 16 slice scanners received higher exam CTDIvol values ranging from 8.6 mgy to 31.2 mgy. These values were too high to qualify as LDCT, especially in regards to AAPM recommendations and CMS requirements. For these exams, the SSDE was larger than the CTDIvol for all 22 patients with effective diameters smaller than 35 cm, and smaller than the CTDIvol for the 1 patient with effective diameter larger than 35 cm. The distribution of assigned Lung-RADS scores was as follows: 64, 34, 12, and 9 patients for Categories 1, 2, 3, and 4, respectively. Cancer was detected in 2 patients with adenocarcinoma. The first was found in a 62 year old male with 35 pack year history of smoking and was characterized as a 10mm solid spiculated solid nodule at screening. The second cancer was detected in a 73 year old female with a 54 pack-year 145

146 history of smoking and was characterized as a 17 mm spiculated nodule at screening. Both patients underwent video assisted thoracoscopic surgery for lobectomy. 7.7 Discussion The measured lung doses of 3.3 mgy for Subject 5, with an exam CTDIvol of 2.1 mgy, and 6.0 mgy for Subject 7, with an exam CTDIvol of 5.3 mgy, were comparable to results reported elsewhere. David Brenner s calculation based on scan techniques reported in literature for LDCT lung cancer screening examinations found that that the lung dose from a single LDCT exam is between 2.5 mgy and 9.0 mgy (102). Another study used the CT-Expo software product to calculate the mean lung dose of the NLST participants to be about 4.7 mgy, based on an exam with a CTDIvol of 2.9 mgy (103). Another study used the National Cancer Institute s CT dosimetry system (104) to calculate organ doses for 23,773 CT scans involved in the NLST and found the average lung dose to be 4.6 mgy (105). One significant difference between these dosimetry methods is that the direct organ dose measurement performed in this work utilized tube current modulation technology, which has an impact on dose reduction, whereas the participants in the NLST were not scanned with tube current modulation. This work confirmed that utilizing protocols with 120 kvp and 25 SD, 100 kvp and 20 SD, or 100 kvp and 25 SD produces images of acceptable image quality without compromising fine lung detail or the diagnostic confidence to assess lung cancer screening. Furthermore, direct organ dose measurements verified these protocols allow organ dose savings as high as 38%. The two radiologists confirmed that they supported the idea to utilize a higher target noise level of 25 SD, and if they saw patient exams with poor image quality, they would ask to switch the target noise level back to 20 SD. As a result, the standard LDCT lung cancer screening protocol was modified to utilize 146

147 120 kvp, 25 SD, 0.35 s rotation time, pitch on the Toshiba Aquilion ONE and Toshiba Aquilion PRIME CT scanners to exercise ALARA for our lung cancer screening patients. Furthermore, the AAPM-recommended lung cancer screening protocol for these two Toshiba scanners recently changed the target noise level to 25 SD, further enforcing the protocol change. Multiple thoracic radiologists have read several patient examinations at UF Health with the new scan technique and have not complained about image quality. From assessing the exam techniques and resulting doses in 109 clinical patient examinations, 102 exams were acquired with adequate low-dose levels on the Siemens Sensation 16 and Toshiba Aquilion PRIME CT scanners. Seven exams were acquired with improper levels of dose on the Toshiba Aquilion ONE, Toshiba Aquilion 64, and Toshiba Aquilion 16 CT scanners due to a low target noise level setting of 10 or 15 SD (98). Furthermore, The Toshiba Aquilion 64 and 16 slice scanners do not have iterative reconstruction and are not on the list of recommended Toshiba scanners for lung cancer screening by AAPM (98). Although the Siemens Sensation 16 scanners do not have iterative reconstruction, it is listed on the recommended Siemens scanners for lung cancer screening by AAPM. The Toshiba Aquilion ONE CT scanner is equipped with iterative reconstruction and is included in AAPM s list of recommended Toshiba CT scanners for lung cancer screening, with a recommended target noise level of 20 SD during the year of 2015, and 25 SD during The unsuitable scanner and technique utilization has been addressed to CT management to prevent repeating high-dose lung cancer screening exams in the future. 147

148 Screening the initial cohort of patients for lung cancer screening using LDCT at UF Health resulted in the diagnosis of 2 lung cancers, corresponding to a detection rate of 1.7%, compared to the NLST detection rate of 3.8% (96). Negative exams included those assigned a Lung-RADS score of 1 or 2, facilitating return to annual screening. Patients with a positive screening exam, with Lung-RADS of 3 or 4, are managed by the referring physician, who is contacted by the interpreting radiologist for all exams assigned a Lung-RADS score of 4. Disadvantages of annual lung cancer screening include high rate of false positive results, over-diagnosis of cancers that never would have become symptomatic, exposure to ionizing radiation, and cost. In the NLST participants, 96.4% of the positive results in the LDCT group and 94.5% of those in the CXR group were false positive results. Regarding risks from radiation exposure, studies have shown that the effective dose for a single lung cancer screening LDCT examination is about 1.4 msv for a man and 2.4 msv for a female, due to the breast dose (103). This effective dose is higher than another common annual screening exam, mammography, with an effective dose of about 0.4 msv. Another study suggested that the risk of radiation-induced lung cancer associated with repeated LDCT lung cancer screening in smokers may not be negligible (102). Based on an average lung dose of 5.2 mgy, the study calculated that a single LDCT lung cancer screening exam would result in a 0.06% increase of radiationinduced lung cancer, and yearly screening from age 50 would add about 0.85% to the 16.9% lung cancer risk faced by a 50-year-old female smoker (102). With screening still fairly new, the reduction in lung-cancer mortality must be weighed against the potential harms from lung cancer screening. 148

149 Table 7-1. Organ doses measured in Subject 5 scanned with a LDCT LCS protocol Organ Dose (mgy) Organ Avg Max SD Skin Thyroid Breast Lung Liver Stomach SI Colon Ovary Uterus Note Organ doses are reported as the average (Avg), maximum (Max) and standard deviation (SD) of all organ dose measurements acquired in each organ. Organs that only had one measurement do not have a reported SD. Table 7-2. Organ doses measured in Subject 7 scanned with a LDCT LCS protocol Organ Dose (mgy) Organ Avg Max SD Skin Thyroid Breast Lung Liver Stomach SI Colon Ovary Uterus Note Organ doses are reported as the average (Avg), maximum (Max) and standard deviation (SD) of all organ dose measurements acquired in each organ. Organs that only had one measurement do not have a reported SD. 149

150 Organ Dose (mgy) Thyroid Liver Skin Breast Lung Stomach Colon Ovary SI Uterus Subject 5 Subject 7 Figure 7-1. Average organ doses measured in Subject 5 and Subject 7 scanned with LDCT LCS protocols 150

151 A B Figure 7-2. Axial chest images reconstructed with the Standard Body Axial algorithm at (A) 1-mm and (B) 3-mm slice thicknesses. Figure courtesy of author. 151

152 A B Figure 7-3. Axial lung images reconstructed with the Lung Sharp algorithm at (A) 1-mm and (B) 3-mm slice thicknesses. Figure courtesy of author. 152

153 A B C Figure 7-4. Axial lung images reconstructed with (A) Lung Sharp, (B) Lung Low Dose and (C) Lung HRCT reconstruction algorithms. Figure courtesy of author. 153

154 Table 7-3. Scan techniques utilized on Subject 3 scanned with various LDCT Lung Cancer Screening protocols for image quality assessment Scan Protocol 1 2 Tube Voltage (kvp) Slice thickness (mm) Number of detectors Rotation time (s) Pitch Focal spot size Small Small Field of View (mm) Bow-tie filter L L Scan Range (mm) Target noise level (SD) Effective mas (mas) ma Range CTDIvol (mgy) DLP (mgy-cm) Table 7-4. Scan techniques utilized on Subject 4 scanned with various LDCT Lung Cancer Screening protocols for image quality assessment Scan Protocol Tube Voltage (kvp) Slice thickness (mm) Number of detectors Rotation time (s) Pitch Focal spot size Small Small Small Small Field of View (mm) Bow-tie filter L L L L Scan Range (mm) Target noise level (SD) Effective mas (mas) ma Range CTDIvol (mgy) DLP (mgy-cm)

155 Table 7-5. Scan techniques utilized on Subject 5 scanned with various LDCT Lung Cancer Screening protocols for image quality assessment Scan Protocol Tube Voltage (kvp) Slice thickness (mm) Number of detectors Rotation time (s) Pitch Focal spot size Small Small Small Small Small Small Field of View (mm) Bow-tie filter L L L L L L Scan Range (mm) Target noise level (SD) Effective mas (mas) ma Range CTDIvol (mgy) DLP (mgy-cm) Table 7-6. Scan techniques utilized on Subject 7 scanned with various LDCT Lung Cancer Screening protocols for image quality assessment Scan Protocol Tube Voltage (kvp) Slice thickness (mm) Number of detectors Rotation time (s) Pitch Focal spot size Large Large Large Large Large Large Large Large Field of View (mm) Bow-tie filter LL LL LL LL LL LL LL LL Scan Range (mm) Target noise level (SD) Effective mas (mas) ma Range CTDIvol (mgy) DLP (mgy-cm)

156 Table 7-7. Scan techniques utilized on Subject 8 scanned with various LDCT Lung Cancer Screening protocols for image quality assessment Scan Protocol Tube Voltage (kvp) Slice thickness (mm) Number of detectors Rotation time (s) Pitch Focal spot size Small Small Small Small Small Small Small Small Field of View (mm) Bow-tie filter L L L L L L L L Scan Range (mm) Target noise level (SD) Effective mas (mas) ma Range CTDIvol (mgy) DLP (mgy-cm) Table 7-8. Scan techniques utilized on Subject 9 scanned with various LDCT Lung Cancer Screening protocols for image quality assessment Scan Protocol Tube Voltage (kvp) Slice thickness (mm) Number of detectors Rotation time (s) Pitch Focal spot size Small Small Small Small Small Small Small Small Field of View (mm) Bow-tie filter L L L L L L L L Scan Range (mm) Target noise level (SD) Effective mas (mas) ma Range CTDIvol (mgy) DLP (mgy-cm)

157 Table 7-9. Scan techniques utilized on Subject 10 scanned with various LDCT Lung Cancer Screening protocols for image quality assessment Scan Protocol Tube Voltage (kvp) Slice thickness (mm) Number of detectors Rotation time (s) Pitch Focal spot size Small Small Small Small Small Small Small Small Field of View (mm) Bow-tie filter L L L L L L L L Scan Range (mm) Target noise level (SD) Effective mas (mas) ma Range CTDIvol (mgy) DLP (mgy-cm) Table Scan techniques utilized on Subject 11 scanned with various LDCT Lung Cancer Screening protocols for image quality assessment Scan Protocol Tube Voltage (kvp) Slice thickness (mm) Number of detectors Rotation time (s) Pitch Focal spot size Small Small Small Small Small Small Small Small Field of View (mm) Bow-tie filter L L L L L L L L Scan Range (mm) Target noise level (SD) Effective mas (mas) ma Range CTDIvol (mgy) DLP (mgy-cm)

158 Table Radiologists utilized to assess image quality of lung cancer screening scans Radiologist ID Radiologist Specialty Radiologist Name 1 Thoracic Nupur Verma 2 Thoracic Tan-Lucien Mohammed Table Image quality scores for diagnosing lung cancer screening in Subject 3 Scan Protocol Image quality feature Radiologist ID Sharp reproduction of lung detail Confidence in diagnosing nodules Confidence in diagnosing lung disease Table Image quality scores for diagnosing lung cancer screening in Subject 4 Scan Protocol Image quality feature Radiologist ID Sharp reproduction of lung detail Confidence in diagnosing nodules Confidence in diagnosing lung disease

159 A B Figure 7-5. Lung cancer screening CT images of Subject 5 acquired at 120 kvp with the protocol using the A) highest dose and B) lowest dose. Figure courtesy of author. 159

160 Table Image quality scores for diagnosing lung cancer screening in Subject 5 Scan Protocol Image quality feature Radiologist ID Sharp reproduction of lung detail Confidence in diagnosing nodules Confidence in diagnosing lung disease Table Image quality scores for diagnosing lung cancer screening in Subject 7 Image quality feature Sharp reproduction of lung detail Radiologist ID Scan Protocol Confidence in diagnosing nodules Confidence in diagnosing lung disease

161 A B Figure 7-6. Lung cancer screening CT images of Subject 8 with the protocol using the A) highest dose and B) lowest dose. A 3-mm granuloma is present in the right lobe. Figure courtesy of author. 161

162 Table Image quality scores for diagnosing lung cancer screening in Subject 8 Image quality feature Sharp reproduction of lung detail Radiologist ID Scan Protocol Confidence in diagnosing nodules Visualization of an existing nodule Confidence in diagnosing lung disease

163 A B Figure 7-7. Lung cancer screening CT images of Subject 9 with the protocol using the A) highest dose and B) lowest dose. An 18-mm speculated nodule is present in the right lobe. Figure courtesy of author. 163

164 Table Image quality scores for diagnosing lung cancer screening in Subject 9 Image quality feature Sharp reproduction of lung detail Radiologist ID Scan Protocol Confidence in diagnosing nodules Visualization of an existing nodule Confidence in diagnosing lung disease Table Image quality scores for diagnosing lung cancer screening in Subject 10 Image quality feature Sharp reproduction of lung detail Radiologist ID Scan Protocol Confidence in diagnosing nodules Visualization of an existing nodule Confidence in diagnosing lung disease

165 VGAS Table Image quality scores for diagnosing lung cancer screening in Subject 11 Image quality feature Sharp reproduction of lung detail Radiologist ID Scan Protocol Confidence in diagnosing nodules Visualization of an existing nodule CT Protocol Number Radiologist 1 Radiologist 2 Figure 7-8. Image quality scores for diagnosing lung cancer screening with 8 CT protocols. Image quality scores were averaged over all postmortem subjects scanned with lung cancer screening CT. VGAS are shown for two radiologists and 8 scan protocols. 165

166 Table Scan techniques utilized on Subject 5 scanned with various Low-Dose Lung Cancer Screening CT protocols for organ dose measurement Scan Protocol Tube Voltage (kvp) Slice thickness (mm) Number of detectors Rotation time (s) Pitch Focal spot size Small Small Small Small Field of View (mm) Bow-tie filter L L L L Scan Range (mm) Target noise level (SD) Effective mas (mas) ma Range CTDIvol (mgy) DLP (mgy-cm) Table Scan techniques utilized on Subject 7 scanned with various Low-Dose Lung Cancer Screening CT protocols for organ dose measurement Scan Protocol Tube Voltage (kvp) Slice thickness (mm) Number of detectors Rotation time (s) Pitch Focal spot size Large Large Large Large Field of View (mm) Bow-tie filter LL LL LL LL Scan Range (mm) Target noise level (SD) Effective mas (mas) ma Range CTDIvol (mgy) DLP (mgy-cm)

167 Table Organ doses measured in Subject 5 scanned with standard and reduceddose LDCT LCS protocols that produced acceptable image quality Organ Organ Dose (mgy) Protocol 1 Protocol 2 Protocol 4 Protocol 5 Avg Max SD Avg Max SD Avg Max SD Avg Max SD Skin Thyroid Breast Lung Liver Stomach SI Colon Ovary Uterus Note: The standard protocol is protocol 1, and the reduced-dose protocols are protocols 2, 4, and 5. Organ doses are reported as the average (Avg), maximum (Max) and standard deviation (SD) of all organ dose measurements acquired in each organ. Organs that only had one measurement do not have a reported SD. 167

168 Table Organ doses measured in Subject 7 scanned with standard and reduceddose LDCT LCS protocols that produced acceptable image quality Organ Organ Dose (mgy) Protocol 1 Protocol 2 Protocol 5 Protocol 6 Avg Max SD Avg Max SD Avg Max SD Avg Max SD Skin Thyroid Breast Lung Liver Stomach SI Colon Ovary Uterus Note: The standard protocol is protocol 1, and the reduced-dose protocols are protocols 2, 5, and 6. Organ doses are reported as the average (Avg), maximum (Max) and standard deviation (SD) of all organ dose measurements acquired in each organ. Organs that only had one measurement do not have a reported SD. 168

169 Organ Dose (mgy) Liver Thyroid Breast Skin Lung Stomach Colon SI Ovary Uterus Protocol Number Figure 7-9. Average organ doses measured in Subject 5 scanned with LDCT lung cancer screening protocols that produced acceptable image quality. Protocols are described in Table The standard protocol is protocol 1, and the reduced-dose protocols are protocols 2, 4, and

170 Organ Dose (mgy) Thyroid Liver Skin Breast Stomach Lung SI Colon Ovary Uterus Protocol Number Figure Average organ doses measured in Subject 7 scanned with LDCT lung cancer screening protocols that produced acceptable image quality. Protocols are described in Table The standard protocol is protocol 1, and the reduced-dose protocols are protocols 2, 5, and

171 Table Percent organ dose reductions achieved in Subject 15 with reduced-dose LDCT LCS protocols that produced acceptable image quality Organ Percent Dose Reduction (%) Protocol 2 Protocol 4 Protocol 6 Skin 23% 13% 21% Thyroid 33% 19% 31% Breast 37% 16% 23% Lung 27% 8% 20% Liver 24% 22% 26% Stomach 34% 2% 25% Average 30% 14% 24% Percent organ dose reductions were calculated against the standard scan (protocol 1) Table Percent organ dose reductions achieved in Subject 17 with reduced-dose LDCT LCS protocols that produced acceptable image quality Organ Percent Dose Reduction (%) Protocol 2 Protocol 5 Protocol 6 Skin 15% 35% 32% Thyroid 30% 26% 38% Breast 27% 36% 35% Lung 21% 36% 37% Liver 17% 33% 34% Stomach 12% 38% 35% Average 20% 34% 35% Percent organ dose reductions were calculated against the standard scan (protocol 1) 171

172 Figure Plot of tube current modulating throughout the scan length of Subject 5 in the anteroposterior (left) and lateral directions (right). The tube current is displayed in yellow for protocol 1, orange for protocol 2, green for protocol 4, and blue for protocol 5. Figure courtesy of author. Figure Plot of tube current modulating throughout the scan length of Subject 7 in the anteroposterior (left) and lateral directions (right). The tube current is displayed in yellow for protocol 1, orange for protocol 2, green for protocol 5, and blue for protocol 6. Figure courtesy of author. 172

173 BMI (kg/m2) Effective Diameter (cm) Figure Histogram of patient weight for 109 patients examined with LDCT lung cancer screening examinations at UF Health 50 R² = R² = Weight (lb) Weight (lb) Figure Correlation of patient weight with BMI (left) and effective diameter in the chest (right) 173

174 Dose (mgy) CTDIvol SSDE Patient Effective Diameter (cm) Figure CTDIvol and SSDE for 86 LDCT lung cancer screening patient examinations conducted on Siemens Sensation 16 CT Scanners at UF Health as a function of patient effective diameter 174

175 Dose (mgy) CTDIvol SSDE Patient Effective Diameter (cm) Figure CTDIvol and SSDE for 23 LDCT lung cancer screening patient examinations conducted on four Toshiba CT Scanners at UF Health as a function of patient effective diameter 175

176 CHAPTER 8 PROTOCOL OPTIMIZATION IN PEDIATRIC BODY CT 8.1 Pediatric CT Imaging With its unmatched ability to provide rapid and detailed information about internal anatomy, computed tomography (CT) has become a valuable tool for diagnosing illness and injury in children (106). The introduction of helical multidetector CT scanners has allowed for faster scanning speeds and improved image quality, which are especially beneficial for examining children. Scans that once required several minutes can now be completed in a matter of seconds, eliminating the need for sedation and making more types of CT examinations available for younger or less cooperative children (107). Following these advancements, the use of CT in children increased dramatically in the past two decades (4,108). In 2011, 85 million CT exams were performed in the United States, with 5% to 11% of these scans performed on children (4,109). This translates to approximately 4.2 to 9.4 million CT examinations performed on children in the United States. A study conducted by Miglioretti et al reported that from 1996 to 2007, the use of CT doubled in children younger than 5 years old, and tripled in children between 5 to 14 years of age (14). Although CT provides diagnostic benefits to pediatric patients, there are concerns about risks associated with radiation exposure. While radiation exposure is a concern in both adults and children, the risk is more significant in children. Owing to the increased number of dividing cells in growing children, children are more sensitive to radiation-induced cancers than adults. Furthermore, children have a longer life expectancy than adults, with more years left for a potential radiation-induced cancer to develop (4,110,111). Additionally, children may receive unnecessarily high radiation 176

177 dose if the CT scan technique settings are not adjusted for their smaller body size. As a result, the risk of developing a radiation-induced cancer can be several times higher. Many methods have been introduced to minimize radiation exposure to children. The purpose of this investigation was to measure typical organ doses to children resulting from standard pediatric CT exams, quantify organ dose reduction from currently used dose reduction methods, and explore the possibility for acquiring diagnostic-quality images with lower dose levels. 8.2 Validation of a Postmortem Subject as a Surrogate for a Pediatric Child It was not possible to obtain a pediatric postmortem subject for this investigation due to the fact that two local suppliers of postmortem subjects do not provide subjects under the age of 18. Therefore, a small-sized adult postmortem subject, Subject 5, was selected as a surrogate for a pediatric patient. The first part of this chapter discusses how we validated the small cadaver was a suitable surrogate for a pediatric patient Methods The subject s measured height and weight values were compared against national averages reported by the National Center for Health Statistics (112), and the subject s effective diameter was compared against averages reported by AAPM Report TG 204 (51). To determine whether the internal anatomy matched that of a pediatric patient, three pediatric radiologists visually compared its intraabdominal fat and individual organs to pediatric patients. The reason for evaluating the intraabdominal fat is because pediatric scan protocols use a lower noise index level than adult protocols due to the decreased intraabdominal fat in children affecting the contrast to noise ratio. 177

178 8.2.2 Results Based on its height of 4 4, the subject corresponded to an average 9-year-old female, and based on its weight of 87 lbs, the subject corresponded to an average year-old female (112). The subject s effective diameter was measured to be 22.3 cm in the central slice of the abdomen scan, corresponding to an average 11-year-old patient (51). Based on its weight, height, and effective diameter, the subject s external metrics corresponded to a female between the ages of 9 and 11.6 years old. For simplicity, the subject will be compared to the size of a 10-year-old female. When visually comparing the subject s intraabdominal fat to a typical 10-year-old child, the radiologists had varying opinions. Two of the radiologists stated that although the subject was very thin, she had more intraabdominal fat than a typical 10-year-old child. They stated that 10-year-old patients have essentially no intraabdominal fat, and their organs are tightly packed together. However, one radiologist stated that the subject had nearly no fat, and she would normally expect more weight and fat in a typical 10- year-old child. The radiologists established that the thyroid, breast, liver, and skin were comparable in size and location to a 10-year-old. They stated she had essentially no breasts, and breast sizes do vary in children. The lung tissue volume and ovaries were larger than a 10-year-old patient. Based on these findings, the following work only measured organ doses for the thyroid, breasts, liver, and skin. 8.3 Measuring Organ Doses from Pediatric Body CT In order to properly assess risk to children, there is a need for knowledge of accurate organ doses arising from examinations performed on modern CT scanners that have advanced dose reduction capabilities. The purpose of this investigation was to 178

179 measure typical organ doses for a 10-year-old pediatric patient being scanned with typical chest and abdomen CT scans on a Toshiba Aquilion ONE CT scanner Methods Subject 5 was scanned with a standard pediatric chest CT protocol and a standard pediatric abdomen CT protocol with scan parameters listed in Table 8-1. Both protocols utilized nearly the same scan techniques with 100 kvp, tube current modulation and iterative reconstruction utilized with a target noise level of 10 SD, with the scan range from above the lung apices to below the adrenal glands for the chest scan and from the top of the hemidiaphragm to below the ischium for the abdomen scan. Organ doses were measured with 2 OSLDs in the thyroid, 8 in the breasts, 5 in the liver, and 15 distributed across the skin Results The chest protocol had an exam CTDIvol of 3.8 mgy and the abdomen protocol had an exam CTDIvol of 4.2 mgy. Organ doses resulting from the pediatric chest scan and pediatric abdomen scan are listed in Table 8-2. For the chest scan, the thyroid received the highest dose at 6.18 mgy, and for the abdomen scan, the liver received the highest dose at 5.36 mgy. Average organ doses for the standard pediatric chest CT protocol and a standard pediatric abdomen CT protocol are graphed in Figure Measuring Organ Doses from Utilizing Current Dose Reduction Methods A major effort has been undertaken to enforce the ALARA principle, utilize CT imaging protocols that account for patient age and size, and apply dose reduction systems. It is therefore of interest to quantify how such dose reduction strategies affect organ doses. 179

180 8.4.1 Appropriateness of CT Examination To minimize radiation dose in pediatric CT, the first step should be to consider alternative imaging modalities that do not utilize ionizing radiation, such as ultrasound or magnetic resonance imaging (MRI), that can properly answer the clinical question. Communication between physicians and radiologists is necessary to determine the need for CT imaging. Furthermore, radiologists should review the indications prior to each pediatric scan, given that a recent publication showed that referring clinicians have little knowledge of the radiation dose or risk that patients are exposed to during CT examinations (113) Child-Sized Techniques If a CT examination is considered necessary for a child, the scan protocol utilized must be designed for children. In 2001, scientific articles received media attention when they demonstrated that using inappropriate adult CT protocols for pediatric patients resulted in unnecessarily high radiation dose to children (110,114). Since then, adapting the dose level to patient size has become a common practice in the CT community, which is further endorsed by the requirement in CT ACR accreditation. Furthermore, the message of the Image Gently campaign to reduce or child-size the amount of radiation used in children encourages hospitals to adopt pediatric CT protocols that adjust exposure parameters for pediatric CT based on child size and weight (115). Centers that receive low numbers of pediatric referrals may lack familiarity with pediatric procedures and may fail to optimize the exam parameters, increasing the dose to the pediatric patient. Patient size-dependent protocols include the use of automatic tube current modulation (TCM), manual technique charts, and bowtie filters. The tube current is 180

181 modulated based on patient size, with the adjustment driven by target noise levels. Generally, for pediatrics, a lower target noise level is demanded owing to reduced natural contrast arising from a lower amount of intraabdominal fat in children compared to adults. For this reason, the pediatric abdomen CT protocol and pediatric chest CT protocol at UF Health utilizes a tube voltage of 100 kvp and a target noise level of 10 SD, in contrast with the adult body protocols that utilize 120 kvp and 12.5 SD Iterative Reconstruction During the last few years, iterative reconstruction algorithms have found great popularity in the clinical community due to their dose reduction potentials. Published data indicate that iterative reconstruction has the potential to significantly reduce the radiation dose up to 76 % while preserving image quality (29). With iterative reconstruction becoming the routine technique of image reconstruction, it has the potential to obtain low-dose CT protocols in children at an acceptable noise level Organ shielding The use of bismuth shields is another way of reducing CT dose to radiosensitive superficial tissues such as the breast, thyroid, testes, and lens of the eye. Commercially available, inexpensive and easy to use, the bismuth shield can offer up to 57% dose reduction to the breast depending on total shield thickness used (37). The use of bismuth breast shields is an implemented policy of the Department of Radiology at UF Health Methods To measure the dose reduction achieved by using an adequate pediatric protocol, organ doses were measured with a pediatric abdomen CT protocol and an adult abdomen CT protocol with techniques listed in Table 8-3. The main differences between the protocols were the tube voltage of 100 kvp for pediatric and 120 kvp for 181

182 adult protocols, the target noise level of 10 SD for pediatric and 12 SD for adult protocols, and TCM algorithms tailored to pediatric and adult populations. Both protocols utilized iterative reconstruction, as per routine clinical practice on the Toshiba Aquilion ONE scanner. To measure the dose reduction achieved by using Toshiba s iterative reconstruction algorithm, the subject was scanned with a pediatric abdomen protocol with filtered back projection and with a pediatric abdomen protocol with iterative reconstruction, with techniques listed in Table 8-4. Breast doses were measured with a pediatric chest protocol performed with and without the placement of a pediatric breast bismuth shield, with techniques listed in Table 8-5. The scan techniques were identical for both scans. The bismuth shield was applied after the localizer scanogram image was obtained to prevent the TCM from increasing the tube current due to the added shield attenuation Results The exam CTDIvol values were 3.7 mgy for the adult protocol and 4.2 mgy for the pediatric protocol. Table 8-6 shows the measured organ doses and percent dose reduction achieved in the abdomen scans by utilizing pediatric protocols. By using an appropriate pediatric protocol with correct techniques, dose reduction of up to 16.7% was realized. This demonstrates the importance of sized-right-imaging. Being that this is a very thin subject, dose savings are expected to be greater for a larger sized pediatric patient. Additionally, because the advanced scanner uses TCM for both pediatric and adult protocols, it was utilized in both scenarios. However, the adult TCM recognizes the small size of the subject and reduces the tube current accordingly. If the same subject was scanned on an older scanner without adjusting for patient size, the dose difference 182

183 between the incorrect adult protocol and correct pediatric protocol would be more prominent. The exam CTDIvol values were 8.1 mgy for filtered backprojection protocol and 4.2 mgy for the iterative reconstruction protocol. Table 8-7 shows the measured organ doses and percent dose reduction achieved in the abdomen scans by utilizing iterative reconstruction. Iterative reconstruction offered the greatest dose reduction up to 54.6%. This stresses that iterative reconstruction is the new path for the image gently campaign, and if possible, all pediatric patients should be routed to a CT scanner with iterative reconstruction. Both chest protocols utilized the same scan settings and exam CTDIvol of 3.8 mgy. Table 8-8 shows the measured breast doses and percent dose reduction achieved in the pediatric chest scans by utilizing bismuth breast shields. The breast shield resulted in 21% dose reduction to the breasts, which is important for young girls with radiosensitive developing breast tissue. Implementation of these readily available dose reduction strategies, combined with the elimination of unnecessary imaging, could dramatically reduce increased risk of stochastic effects from CT use in pediatrics. 8.5 Assessing Image Quality for Reduced-Dose Pediatric CT Exams Given that children are two to ten times more radiosensitive than adults (95,96), it remains essential to apply the ALARA principle in pediatric CT examinations. The BEIR VII report estimated that an exposure of 10 msv carries a 1 in 1000 risk of developing a solid cancer or leukemia (7). The benefits of properly performed and clinically justified CT examinations should always outweigh the risks for an individual child; where unnecessary exposure is associated with unnecessary risk. For some exams, the 183

184 highest quality images are not always necessary to make a diagnosis, allowing for dose to be reduced for certain applications that have higher intrinsic contrast. For chest CT, the high inherent contrast and low attenuation in lungs allow tolerance of image noise and dose reduction compared with other body regions, such as the abdomen or head. Abdominal CT, on the other hand, is more challenging for dose optimization because of organs with low contrast enhancement, such as the liver. Diagnostic quality depends on radiologist preference and clinical indication, making it a complex challenge to identify the optimal balance of diagnostic quality and radiation dose. The next part of this investigation attempts to identify at what dose level is diagnostic quality impaired for pediatric chest and abdomen scans at our institution Methods To assess image quality at lower dose levels, pediatric chest and pediatric abdomen CT scans were acquired using five target noise levels ranging from the standard 10 SD up to 35 SD. The five scan protocols are listed in Table 8-9 for the chest scans and Table 8-10 for the abdomen scans. Using the image quality categories described in Chapter 5, three pediatric radiologists listed in Table 8-11 graded the five scans with a three-point scale Results The reduction of CTDIvol values with increasing target noise level is represented in Figure 8-2. The image quality scores obtained from the blinded observer study are displayed in Table 8-12 for the chest scans and Table 8-13 for the abdomen scans. Perfect inter-observer agreement was found among the three radiologists for the chest scans acquired with standard-dose (protocol 1) and reduced-dose (protocols 2 and 3), with an intra-class correlation coefficient of 1.00 (p = 0) for the VGAS. All 184

185 radiologists scored all 6 image criteria to be of acceptable image quality. For the reduced-dose protocol number 4, almost perfect inter-observer agreement among the participating radiologists was found by means of an intra-class correlation coefficient of 0.86 (p < 0.001) for the VGAS. For the reduced-dose protocol number 5, moderate inter-observer agreement among the participating radiologists was found by means of an intra-class correlation coefficient of 0.56 (p > 0.001) for the VGAS. Since all protocols produced at least moderate inter-observer agreement, image quality scores were averaged over all readers and displayed in Figure 8-3. Perfect inter-observer agreement was found among the three radiologists for the chest scans acquired with standard-dose (protocol 1) and reduced-dose (protocol 2), with an intra-class correlation coefficient of 1.00 (p = 0) for the VGAS. All radiologists scored all 6 image criteria to be of acceptable image quality. All radiologists scored all 6 image criteria to be of acceptable image quality. For the reduced-dose protocol number 3, moderate inter-observer agreement among the participating radiologists was found by means of an intra-class correlation coefficient of 0.53 (p > 0.001) for the VGAS. For the reduced-dose protocol number 4, moderate inter-observer agreement among the participating radiologists was found by means of an intra-class correlation coefficient of 0.53 (p > 0.001) for the VGAS. For the reduced-dose protocol number 5, substantial inter-observer agreement among the participating radiologists was found by means of an intra-class correlation coefficient of (p < 0.001) for the VGAS. Since all protocols produced at least moderate inter-observer agreement, image quality scores were averaged over all readers and displayed in Figure

186 The chest scans were found to be of acceptable image quality up to a target noise level of 17.5 SD, and abdomen scans were acceptable up to a target noise level of 12.5 SD. All other protocols resulted in image quality features scored as borderline acceptable or unacceptable. 8.6 Measuring Organ Doses from Reduced-Dose Pediatric CT Exams Methods Organ doses were measured for protocols that produced acceptable scores, which were protocols with 10 SD, 12.5 SD, and 17.5 SD for chest CT protocols and 10 SD and 10 SD and 12.5 SD for abdomen CT protocols Results Measured organ doses are shown in Table 8-14 for the chest protocols and Table 8-15 for the abdomen protocols. The percent dose reductions are listed in Table 8-16 for each organ. Organ doses reduced by 44-52% in the chest and 17-47% in the abdomen, while still producing diagnostic image quality. 8.7 Discussion Miglioretti et al (14) recorded clinical CT exam parameters of 147 pediatric CT chest CT exams conducted at 4 large hospital systems and estimated organ doses using organ dose conversion coefficients based on ICRP computational anatomy phantoms and Monte Carlo simulations (116). Their calculated mean organ doses for a patient between the ages of 10 to 14 years old are listed in Table Miglioretti s reported average organ doses were higher than our measured organ doses. The study extracted dose data from 5 hospital systems with various CT scanners, without reporting the scanner models and variation in techniques. Averaging doses from scanners that utilized filtered backprojection and iterative reconstruction may not 186

187 provide an accurate estimate of a patient who is scanned on a scanner equipped with iterative reconstruction. Because our measured organ doses were not an average of many exam protocols, but were measured in one subject scanned with one protocol on one scanner, it is expected for the dose to be lower than Miglioretti s average as our protocol exclusively used iterative reconstruction. Using the organ dose conversion coefficients by Lee et al (116), we calculated organ doses for Subject 5. For the chest CT scan, the effective mas reported by the scanner was 116 mas. To convert effective mas to mean mas, the effective mas of 116 mas was multiplied by the pitch of to give a mean tube-current-time-product of 96 mas. Organ doses were calculated with an exam CTDIvol of 3.8 mgy and 96 mas, and are listed in Table The calculated organ dose for Subject 5 using Lee s conversion coefficients resulted in larger organ dose values. However, the organ dose conversion coefficients do not include the effect of tube current modulation, and requires the user to correct for slice-specific mas. The authors of the computational method (104) stated that when using the mean mas, dose to the organs close to the lungs will be significantly overestimated in chest and CAP scans. For example, they stated that breast doses are 1.7 times larger for a scan that did not modulate ma. This was indeed true, where Lee s method estimated breast dose of 7.01 mgy for a 10-year-old child, and the direct measured breast dose with tube current modulation was in fact 5.32 mgy, 1.3 times smaller than calculated. To analyze how the mean tube current compares to the modulating tube current, the mean tube current time product was divided by the rotation time of 0.5 s to give a mean tube current of 192 ma. Figure 8-5 shows the modulating tube current in blue in 187

188 the anteroposterior and lateral planes of Subject 5, as well as the mean tube current of 192 ma shown in red. It can be seen that the thyroid also received less tube current than the average 192 ma. As a result, the thyroid dose calculated by Lee s method with a mean tube current was overestimated by 1.34 times larger than when directly measured with a modulating tube current. When using bismuth shielding, our measured breast dose reduction of 21% is comparable to results reported by Coursey et al (35), who stated that bismuth breast shield reduced dose to the breast of a 5-year-old anthropomorphic phantom by 26%. Coursey further stated that when compared to a scan without modulating tube current, the use of breast shield with automatic tube current modulation reduced the breast dose by 52%. Although we did not measure breast dose with a fixed tube current protocol, it is of interest to note that the utilization of tube current modulation with bismuth breast shields offer significant dose savings to pediatric patients. This work has limitations. Although we report pediatric organ doses, the measurements were not carried out in a pediatric cadaver, but a child-sized cadaver which served as a surrogate. Furthermore, organ doses were only investigated in one cadaver due to the difficulty of acquiring postmortem subjects that are small-sized. Additionally, the greatest concern for radiation risk exists for younger children, including those who are scanned at just a few months or years old. Patients at that age and size cannot be investigated with direct dosimetry methods in cadavers, and the organ doses reported in this work are not applicable to those young patients. Other work in literature addresses organ doses received by young children using computational phantoms (116). 188

189 Table 8-1. Scan techniques utilized on Subject 5 scanned with a standard pediatric chest CT protocol and a standard pediatric abdomen CT protocol Scan Parameters Chest Protocol Abdomen Protocol Tube Voltage (kvp) Slice thickness (mm) Number of detectors Rotation time (s) Pitch Focal spot size Small Small Field of View (mm) Bow-tie filter M M Anatomical Scan Range Above the lung apices to below the adrenal glands Scan Range (mm) Target noise level (SD) Effective mas (mas) Reconstruction AIDR-3D AIDR-3D CTDIvol (mgy) DLP (mgy-cm) Top of the hemidiaphragm to below the ischium Table 8-2. Organ doses measured in Subject 5 scanned with a standard pediatric chest CT protocol and a standard pediatric abdomen CT protocol Organ Dose (mgy) Organ Chest Protocol Abdomen Protocol Avg Max SD Avg Max SD Thyroid Breast Liver Skin Organ doses are reported as the average (Avg), maximum (Max) and standard deviation (SD) of all organ dose measurements acquired in each organ. 189

190 Organ Dose (mgy) Chest 3 Abdomen Thyroid Breast Liver Skin Organ Figure 8-1. Average organ doses measured in Subject 5 scanned with a standard pediatric chest CT protocol and a standard pediatric abdomen CT protocol 190

191 Table 8-3. Scan techniques utilized on Subject 5 scanned with abdomen CT protocols using adult settings versus pediatric settings Scan Parameters Adult Pediatric Tube Voltage (kvp) Slice thickness (mm) Number of detectors Rotation time (s) Pitch Focal spot size Small Small Field of View (mm) Bow-tie filter M M Scan Range (mm) Target noise level (SD) Effective mas (mas) Reconstruction AIDR-3D AIDR-3D CTDIvol (mgy) DLP (mgy-cm) Note: Both scans were acquired from top of the hemidiaphragm to below the ischium Table 8-4. Scan techniques utilized on Subject 5 scanned with pediatric abdomen CT protocols acquired with and without iterative reconstruction algorithms Scan Parameters Filtered Backprojection Iterative Reconstruction Tube Voltage (kvp) Slice thickness (mm) Number of detectors Rotation time (s) Pitch Focal spot size Small Small Field of View (mm) Bow-tie filter M M Scan Range (mm) Target noise level (SD) Effective mas (mas) Reconstruction FBP AIDR-3D CTDIvol (mgy) DLP (mgy-cm) Note: Both scans were acquired from top of the hemidiaphragm to below the ischium 191

192 Table 8-5. Scan techniques utilized on Subject 5 scanned with pediatric chest CT protocols with and without breast bismuth shields Scan Parameters Without breast shielding With breast shielding Tube Voltage (kvp) Slice thickness (mm) Number of detectors Rotation time (s) Pitch Focal spot size Small Small Field of View (mm) Bow-tie filter M M Scan Range (mm) Target noise level (SD) Effective mas (mas) Reconstruction AIDR-3D AIDR-3D CTDIvol (mgy) DLP (mgy-cm) Note: Both scans were acquired from above the lung apices to below the adrenal glands 192

193 Table 8-6. Average organ doses and reductions from utilizing adult versus pediatric scan settings in an abdomen CT scan Organ Average Organ Dose (mgy) Adult Pediatric % Dose Reduction Thyroid Breast Liver Skin Table 8-7. Average organ doses and reductions from utilizing iterative reconstruction versus filtered back projection in a pediatric abdomen scan Organ Average Organ Dose (mgy) Filtered Backprojection Iterative Reconstruction % Dose Reduction Thyroid Breast Liver Skin Table 8-8. Average breast dose and reduction from utilizing bismuth breast shields in a pediatric chest scan Organ Average Breast Dose (mgy) Without breast shielding With breast shielding % Dose Reduction Breast

194 Table 8-9. Scan techniques utilized on Subject 5 scanned with standard and reduceddose pediatric chest CT protocols Scan Protocol Tube Voltage (kvp) Slice thickness (mm) Number of detectors Rotation time (s) Pitch Focal spot size Small Small Small Small Small Field of View (mm) Bow-tie filter L L L L L Scan Range (mm) Target noise level (SD) Effective mas (mas) CTDIvol (mgy) DLP (mgy-cm) Table Scan techniques utilized on Subject 5 scanned with standard and reduceddose pediatric abdomen CT protocols Scan Protocol Tube Voltage (kvp) Slice thickness (mm) Number of detectors Rotation time (s) Pitch Focal spot size Small Small Small Small Small Field of View (mm) Bow-tie filter M M M M M Scan Range (mm) Target noise level (SD) Effective mas (mas) CTDIvol (mgy) DLP (mgy-cm)

195 CTDI vol (mgy) 6 5 Chest Protocol Abdomen Protocol Target Noise Level (SD) Figure 8-2. Exam CTDIvol values achieved by increasing the target noise level values in pediatric chest and abdomen protocols. Table Radiologists utilized to assess image quality of pediatric CT scans Radiologist ID Radiologist Specialty Radiologist Name 1 Pediatric Robert Dubuisson 2 Pediatric Jonathan Williams 3 Pediatric Dhanashree Rajderkar 195

196 Table Image quality scores for diagnosing pediatric chest CT protocols Image quality feature Visualization of lung parenchyma R-ID Scan Protocol Confidence in diagnosing bronchiectasis Soft tissue contrast in peripheral muscles Image noise in peripheral muscles Diagnostic confidence in diagnosing chest wall trauma Overall IQ Acceptance Note: Image quality features were assesed on a 3-point grading scale by three radiologists described by their radiologist ID (R-ID) listed in Table A score of 1 represents unnacceptable, 2 represents borderline acceptable, and 3 represents acceptable image quality. The pediatric chest scan protocols are described in Table

197 Table Image quality scores for diagnosing pediatric abdomen CT protocols Image quality feature Soft tissue contrast in major abdominal organs R-ID Scan Protocol Sharpness in the liver Image noise in the liver Diagnostic confidence in diagnosing umbilical hernia Diagnostic confidence in diagnosing nephrolithiasis Overall IQ Acceptance Note: Image quality features were assesed on a 3-point grading scale by three radiologists described by their radiologist ID (R-ID) listed in Table A score of 1 represents unnacceptable, 2 represents borderline acceptable, and 3 represents acceptable image quality. The five pediatric abdomen scan protocols are described in Table

198 VGAS VGAS CT Protocol Number Figure 8-3. Image quality scores for pediatric chest CT protocols. The standard-dose protocol is represented by protocol number 1, and reduced-dose protocols are represented by protocol numbers 2 through 5, with scan parameters described in Table Visual grading analysis scores (VGAS) were averaged across 3 radiologists who read 6 image quality features CT Protocol Number Figure 8-4. Image quality scores for pediatric abdomen CT protocols. The standarddose protocol is represented by protocol number 1, and reduced-dose protocols are represented by protocol numbers 2 through 5, with scan parameters described in Table Visual grading analysis scores (VGAS) were averaged across 3 radiologists who read 6 image quality features. 198

199 Table Organ doses measured in Subject 5 scanned with standard and reduceddose pediatric chest CT protocols that produced acceptable image quality Organ Organ Dose (mgy) Protocol 1 Protocol 2 Protocol 3 Avg Max SD Avg Max SD Avg Max SD Thyroid Liver Breast Skin Note: Standard (protocol 1) and reduced-dose (protocols 2 and 3) protocols are described in Table 8-9. Organ doses are reported as the average (Avg), maximum (Max) and standard deviation (SD) of organ dose measurements in each organ. Table Organ doses measured in Subject 5 scanned with standard and reduceddose pediatric abdomen CT protocols that produced acceptable image quality Organ Organ Dose (mgy) Protocol 1 Protocol 2 Avg Max SD Avg Max SD Thyroid Breast Liver Skin Note: Standard (protocol 1) and reduced-dose (protocol 2) protocols are described in Table Organ doses are reported as the average (Avg), maximum (Max) and standard deviation (SD) of all organ dose measurements acquired in each organ. Table Maximum achievable organ dose reductions (%) from using pediatric chest and abdomen protocols that produced acceptable image quality Organ Chest Abdomen Skin 52% 25% Thyroid 44% 47% Breast 57% 17% Liver 47% 18% Average 50% 27% Note: The lowest dose acceptable chest protocol was Protocol 3 and the lowest dose acceptable abdomen protocol was Protocol 2. Organ dose reductions were calculated relative to the reference protocol, Protocol

200 Table Average pediatric chest CT organ doses measured in this work compared to organ doses calculated elsewhere Organs Organ Dose (mgy) This work Lee Miglioretti a Thyroid Breast Liver Skin N/A a Miglioretti did not calculate average skin dose A B Figure 8-5. Tube current modulation (blue) and average tube current (red) of Subject 5. Tube current was superimposed on the anteroposterior (A) and lateral (B) scanograms. Figure courtesy of author. 200

201 CHAPTER 9 PROTOCOL OPTIMIZATION IN PEDIATRIC HEAD CT Head CT is the most common CT examination in children, accounting for about 75% of all pediatric CT examinations worldwide (117). A study by Miglioretti et al evaluated the use of CT for over 4 million child-years of observation at 6 major hospital systems in the United States. The group found that the head was the most commonly scanned region, accounting for about 10 percent of pediatric CT exams, in contrast with the abdomen and pelvis, which only accounted for about 2 percent of pediatric CT exams (14). While there are risks associated with head CT exams in children, including an increased risk of brain cancer and leukemia (118), the majority of head CT examinations are conducted to evaluate serious issues. In order to justify the clinical benefit of performing a head CT examination on a child, it is essential to have a proper understanding of the associated radiation risks and to enforce the ALARA principle for this radiosensitive population. 9.1 Measuring Organ Doses from Pediatric Head CT The risks associated with radiation exposure is a concern in the head region due to the radiosensitive organs being imaged. It is well known that the lens of the eye, thyroid, salivary glands, and skin are radiosensitive organs, especially in children. Miglioretti et al calculated brain doses from several pediatric head CT examinations, and found the mean brain dose to be 29.8 mgy (14). The authors also found that 14% of the pediatric head CT examinations resulted in brain doses above 50 mgy, which was estimated by Pearce et al to put children at 2.8 greater risk of developing brain cancer (69). With a large number of pediatric patients undergoing head CT, it was 201

202 important to measure typical organ doses resulting from pediatric head CT exams at UF Health Methods Although Postmortem Subject 5 was not a pediatric postmortem subject, it had the weight and height of a typical 10-year old child, as described in Chapter 8. Furthermore, all standard head CT examinations at UF Health are conducted with the same scan techniques for patients older than 8 years old, whether pediatric or adult. One research study by Kleinman et al measured head sizes for 336 children ranging from 0 to 21 years of age, and reported that the head showed accelerated growth from birth to about 2 years of age, and reached a steady plateau up to 21 years of age. Subject 5 had a transverse head measurement of cm, measured at the maximal parietal diameter between the most lateral points of the parietal bones, and an anteroposterior head measurement of cm, measured between the glabella and the opisthocranion. These measurements fit Kleinman s trend with a lower limit size of a 2-year-old patient and an upper limit size of a 16-year-old patient (119). Because we were investigating doses to a patient older than 8 years old, and because we were using an adult head protocol used on patients older than 8 years old, we found Subject 5 to be a suitable surrogate for a child between 8 to 16 years old that would be scanned with an adult head CT protocol at UF Health. One PVC placement tube was inserted into the brain of Subject 5 by an international fellow of neurosurgery, Dr. Vanessa Milanessi, providing access to the third ventricle, as shown in Figure 9-1. Incisions were made at the thyroid and salivary glands, providing access to insert and replace OSLD holders in a reproducible manner. A total of 20 dosimeters were used for each head scan, with 4 in the brain, 2 in the 202

203 thyroid, 2 in the salivary glands, 10 on the surface of the skin, and 2 placed on top of the eyes. Figure 9-2 shows images of the dosimeter placement in the brain. Subject 5 was scanned with three clinical CT protocols of the head: a standard Head CT protocol, a Trauma Head CT protocol, and a Trauma Head and Cervical Spine CT protocol. All scans were acquired by an experienced CT technologist, Brian Cormack. The Head CT protocol is performed for indications such as headache, intracranial hemorrhage, hydrocephalus, brain injury, and seizure. The scan techniques include a helical acquisition from the base of the skull to the skull vertex, without the use of intravenous contrast, 120 kvp, fixed tube current of 270 ma, gantry rotation time of 0.75 sec, helical pitch of 0.66, 0.5 x 32 detector configuration, resulting in a CTDIvol of 70.8 mgy. The Trauma Head CT protocol is performed for trauma to the head. The scan techniques include a helical acquisition from the base of the skull to the skull vertex, without the use of intravenous contrast, 120 kvp, fixed tube current of 280 ma, gantry rotation time of 0.75 sec, helical pitch of 0.66, 0.5 x 64 detector configuration, resulting in a CTDIvol of 71.4 mgy. The Trauma Head and Cervical Spine CT protocol is performed for trauma to the head and cervical spine. The protocol consists of two separate acquisitions: a Trauma Head scan with fixed tube current and a cervical spine scan with modulating tube current. The head scan uses the same parameters as the Trauma Head CT protocol described above. The cervical spine scan uses a helical acquisition from the sella to the T3 vertebra, without the use of intravenous contrast, with scan parameters of 135 kvp, 203

204 a target noise index of 10 SD, iterative reconstruction, gantry rotation time of 0.75 sec, 0.5 mm x 64 detector configuration, and a CTDIvol of 25.5 mgy. In comparing the Head CT protocol and the Trauma Head CT protocol, the Trauma Head CT protocol uses a higher fixed ma than the Head protocol, with 280 ma rather than 270 ma, as well as a larger scan range than the Head protocol, ending at the chin rather than the base of the skull. Figure 9-3 shows the desired scan range for the Head CT protocol and the Trauma Head CT protocol, as defined by UF Health s Department of Radiology. Figure 9-4 shows screenshots of the scan range in the head for the two protocols in Subject 5, as well as a sagittal view of the cervical spine imaged during the Trauma Head and Cervical Spine protocol. All scans were performed with a 15-degree angulation in order to set the scanning plane parallel to the inferior orbitomeatal (IOM) plane to reduce intracranial artifact. Organ doses were measured for the Head CT protocol, the Trauma Head CT protocol, and the Trauma Head and Cervical Spine CT protocol Results The measured organ doses are listed in Table 9-1 and displayed in Figure 9-5. The Trauma Head and Cervical Spine CT protocol resulted in higher organ doses than the Head CT protocol, especially for the lens of the eye and the thyroid. However, doses were similar for the brain, salivary glands, and skin. One organ of concern was the lens of the eye, which received a dose of for the Head protocol and mgy for the Trauma Head and Cervical Spine protocol. The increased dose to the lens from the Trauma Head and Cervical Spine protocol was due to an additional mgy from the higher-ma Trauma Head scan and an additional mgy from the cervical spine scan. Although the cervical spine 204

205 acquisition did not include the lens of the eye in the scan range for Subject 5, the beginning of the scan range started at just 3 cm below the lens. 9.2 Assessing Image Quality for Reduced-Dose Pediatric Head CT Although pediatric patients between the age of 8 and 18 may have similar head sizes compared to adults, their radiosensitivity is not the same. Thus, dose reduction should be investigated for appropriate scan indications in this radiosensitive population who are being scanned with adult dose levels at UF Health. The purpose of this study was to evaluate the possibility of dose reduction in head CT for specific indications through a systematic reduction in radiation dose paired with a clinical image quality analysis specific to the diagnostic task Indications that Warrant Dose Reduction Previous studies have shown that low-dose CT exams of high-contrast structures, such as lungs or bone, do not result in a loss of diagnostic information (120,121). However, it is less clear whether dose reduction is feasible in imaging of lowcontrast intracranial structures. Differentiation of low-contrast structures such as grey matter (GM) and white matter (WM) require an exceptional contrast-to-noise ratio (CNR) for the detection of subtle pathological findings. In certain clinical circumstances, a trade-off between reduced radiation dose and image quality may be acceptable without sacrificing diagnostic accuracy Hydrocephalus Under normal conditions, there exists a delicate balance between the rate at which cerebrospinal fluid (CSF) is produced and absorbed in the brain. Hydrocephalus is a condition that occurs when the balance is disrupted and the rate of CSF absorption is less than the rate of production. Hydrocephalus can occur at any age, but is most 205

206 common in infants and adults age 60 and older (122). According to the National Institute of Neurological Disorders and Stroke (NINDS), hydrocephalus affects approximately 1 in every 500 children (123). For fetal and neonatal imaging, transfontanelle ultrasonography is useful for diagnosing congenital hydrocephalus. For children and adults, a CT exam can illustrate patterns of ventricular enlargement, although an MRI exam may be required to demonstrate periventricular abnormalities or small obstructing lesions. To our knowledge, no studies have investigated the potential for low-dose CT exams for assessing hydrocephalus. Radiologists who diagnose hydrocephalus at UF Health believe the initial diagnostic exam does not require full-dose CT exams Ventriculoperitoneal shunt Hydrocephalus is most commonly treated with a ventriculoperitoneal (VP) shunt surgically inserted in the brain. VP shunt placement is one of the most commonly performed neurosurgical operations in the United States, with more than 125,000 procedures performed annually (124). CT imaging studies evaluate the integrity of the shunt system, assess changes in ventricular size, and identify shunt-related complications such as mechanical failure, infections, obstructions, or the need to replace or lengthen a catheter. Because as many as 71% of all patients experience shunt malfunction during their lifetime (125), and as many as 40% of shunts fail within the first 2 years after placement (126), the potential for repeated head CT studies is high in patients with VP shunts. When evaluating shunt malfunction, radiologists assess VP shunt localization and ventricular size. Previous studies have shown the utility of low-dose CT for VP shunt follow-up. One study lowered the dose for a standard VP shunt protocol on a GE 4-slice CT scanner by reducing the tube current from 220 ma to 80 ma, resulting in a 206

207 63.4% dose reduction (71). Another study lowered the dose for a standard VP-shunt protocol on a Siemens 128-slice scanner by reducing the tube-current-time-product from 350 to 200 mas, resulting in an average 50% dose reduction (127). Both low-dose protocols produced images of acceptable image quality for the evaluation of shunt failure. Although these studies verified that VP-shunt follow-up exams can be diagnosed with lower doses, it remained essential to conduct our own protocol optimization for VP shunt CT examinations specific to the CT scanners and radiologists at UF Health Trauma follow-up CT imaging is highly sensitive for identifying brain injury requiring acute intervention. Initial CT examinations must detect subtle changes in intracranial structures, requiring high contrast to noise ratio. However, follow-up CT examinations are targeted to identify gross morphological changes in malformation, tumors, trauma, or vascular disease (128). Since these indications often involve large anatomical structures with relatively high contrast-enhancing features, such as in assessing bleeding or ventricular size, they carry potential for dose reduction if the radiologists are willing to accept noisier images than a primary diagnostic examination. Low-dose head CT may be appropriate for routine follow-up of known lesions Craniosynostosis Craniosynostosis is the premature fusion of one or more cranial sutures. It occurs in about 2 per 5,000 live births and results in cranial deformity, increased intracranial pressure, and restricted brain growth, possibly impairing brain function (129). Early diagnostic confirmation is vital to ensure proper management of this condition. Threedimensional computed tomography evaluation is utilized for accurate diagnosis, therapy planning, postoperative, and long term follow-up (129,130). 207

208 Aware of the inherent risks of radiation exposure in the pediatric population, studies have shown the utility of low-dose CT for craniosynostosis assessment with acceptable image quality (72, ). Dose reduction methods include using lower kvp, lower fixed ma, tube current modulation, increased noise index setting, and iterative reconstruction. One study reduced the dose from a standard craniosynostosis exam with a CTDIvol of 31 mgy down to 1.41 mgy while still producing acceptable image quality (132). They achieved dose reduction 22 times less than the 2004 European guidelines for MDCT recommended dosimetric reference levels. Another study reduced the dose from the standard craniosynostosis scan protocol with a CTDIvol of 26.1 mgy down to 2.85 mgy while achieving diagnostic image quality for assessing craniosynostosis (133). At UF Health, any patient over the age of 8 years old is evaluated for craniosynostosis with a standard adult non-contrast head CT exam with 120 kvp, fixed tube current of 270 ma, 0.75 s rotation time, 0.5 x 32 detector configuration, resulting in a CTDIvol of 70.8 mgy. With a high CTDIvol of 70.8 mgy, there appears to be room for dose reduction for CT exams acquired for investigating craniosynostosis Methods Subject 5 was scanned with a standard Head CT protocol with 270 ma, followed by reduced tube currents of 250 ma, 200 ma, and 150 ma. Furthermore, because VP shunt CT scans are currently performed using a Stealth Head protocol, Subject 5 was also scanned with a Stealth Head CT protocol with 250 ma, followed by reduced tube currents of 200 ma, 150 ma, and 100 ma. The main difference between the Head CT protocol and Stealth Head CT protocol is the gantry angulation, where the Stealth Head 208

209 protocol does not use any angulation in order to produce images that are compatible with the neurosurgery software. Tube current modulation is used routinely for chest, abdomen, and pelvis CT studies, but is not typically used in head studies (134). This is due to the head not having areas of drastic anatomical changes, and issues with clinical mis-centering of the head. However, if used correctly, allowing the tube current to modulate ensures that patients with smaller heads receive less dose than patients with larger heads. Sure Exposure 3D was utilized with a target noise level of 5 SD, 7.5 SD, and 10 SD for the standard head protocol and stealth head protocol. With tube current modulation utilized, it was also possible to benefit from the use of iterative reconstruction. AIDR-3D was used with a target noise level of 2.5 SD, 5 SD, 7.5 SD, and 10 SD for the standard Head CT protocol, and 5 SD, 7.5 SD, and 10 SD for the Stealth Head CT protocol. As tube current modulation is not typically used, we were concerned about potential issues that may arise in cases where the patient s head was not centered in the gantry. Because the tube current is calculated based on the scanogram, a shift closer to or farther from the x-ray source would result in magnification effects, making the patient s head appear larger or smaller, respectively. Shifting the head closer to the source, or in this case, an anterior table shift, would magnify the head and result in a larger tube current output. To investigate the effects of a mis-centered patient, we scanned the cadaver s head at center, then shifted anteriorly by 4 cm. Clinical image quality was investigated for indication-specific image quality features on a three-point scale with a pediatric neuroradiologist, Dr. Dhanashree Rajderkar, and the Chair of neuroradiology, Dr. Jeffrey Bennet, as described in Table 9-209

210 2. The image quality was assessed using the blinded observer study methodology described in Chapter Results For the Head CT protocol, reducing the fixed tube current from 270 ma to 250 ma, 200 ma, and 150 ma reduced the CTDIvol from 70.8 mgy to 65.6 mgy, 52.5 mgy, and 39.4 mgy, respectively. Using tube current modulation with a target noise index of 5 SD, 7.5 SD, and 10 SD reduced the CTDIvol from 70.8 mgy to 58 mgy, 32.3 mgy, and 20.4 mgy, respectively. Utilizing AIDR-3D with target noise indices of 2.5 SD, 5 SD, 7.5 SD and 10 SD reduced the CTDIvol from 70.8 mgy to 68.1 mgy, 44 mgy, 20.2 mgy, and 11.8 mgy, respectively. The CTDIvol values for all Head CT protocols are illustrated in Figure 9-6. For the Stealth Head CT protocol, reducing the fixed tube current in the from 250 ma to 200 ma, 150 ma, and 100 ma reduced the CTDIvol from 29.6 mgy to 23.7 mgy, 17.7 mgy, and 11.8 mgy, respectively. Using tube current modulation with a target noise index of 5 SD, 7.5 SD, and 10 SD reduced the CTDIvol to 25.1 mgy, 19.1 mgy, and 15 mgy. Utilizing AIDR-3D with target noise indices of 5 SD, 7.5 SD, and 10 SD reduced the CTDIvol to 22.4 mgy, 15.5 mgy, and 10.3 mgy, respectively. The resulting CTDIvol values for all Stealth Head CT protocols are illustrated in Figure 9-7. The largest dose reduction achieved for both protocols was seen with tube current modulation with AIDR-3D at a target noise index of 10 SD, achieving a CTDIvol reduction of 83% for the standard head protocol and 65% for the stealth head protocol. Screenshots of the tube current modulation plots are displayed in Figure 9-8 for the Head CT protocol and in Figure 9-9 for the Stealth Head CT protocol. Screenshots of the tube current modulation plots produced by the scanner with and without iterative 210

211 reconstruction are displayed in Figure 9-10 for the Head CT protocol and in Figure 9-11 for the Stealth Head CT protocol. Figure 9-12 shows all variations of tube current for the head protocol superimposed on one image to visualize the reductions in tube current from the standard fixed tube current to the tube current modulation variations, to the iterative reconstruction variations, with tube current decreasing as target noise index increased. When the head was centered, the CTDIvol was 22.4 mgy. When the head was shifted 4 cm anteriorly, the CTDIvol increased to 25.5 mgy. The tube current modulation plots for the centered and anterior positions are shown in Figure The image quality scores for assessing hydrocephalus, craniosynostosis, trauma follow-up, and VP shunt are listed in Table 9-3, Table 9-4, Table 9-5, and Table 9-6, respectively. For hydrocephalus, Dr. Rajderkar graded all image quality features as acceptable for all scans with the exception of the two lowest dose protocols acquired with iterative reconstruction with a target noise index of 7.5 SD and 10 SD. Dr. Bennett graded all image quality features acceptable for all scan protocols with the exception of the iterative reconstruction with 10 SD. For craniosynostosis, Dr. Rajderkar and Dr. Bennett graded all image quality features as acceptable for all scans. This was the only exam indication that produced acceptable image quality scores across all image quality features and low-dose protocols. For trauma follow-up, Dr. Rajderkar graded all image quality features as acceptable for the standard protocol, the protocol with tube current fixed at 250 ma, the protocol with tube current modulation with 5 SD, and the protocols with iterative reconstruction with 2.5 SD and 5 SD. Dr. Bennet graded all image quality features acceptable for the standard protocol, the protocols with tube current fixed at 211

212 250 ma and 200 ma, the protocols with tube current modulation at 5 and 7.5 SD, and the protocol with iterative reconstruction at 5 SD and 7.5 SD. For VP shunt, Dr. Rajderkar graded all image quality features as acceptable for the standard protocol, the protocol with tube current fixed at 200 ma, the protocol with tube current modulation with 5 SD, and the protocol with iterative reconstruction with 5 SD. Dr. Bennett graded all image quality features acceptable for the standard protocol, the protocols with tube current fixed at 200 ma and 150 ma, the protocol with tube current modulation with 5 SD, and the protocol with iterative reconstruction with 5 SD. 9.3 Measuring Organ Doses from Reduced-Dose Pediatric Head CT Methods Organ doses were measured for protocols that resulted in acceptable image quality for the purposes of quantifying organ dose reduction, rather than quantifying exam CTDIvol reduction. For hydrocephalus, organ doses were measured for a fixed tube current of 150 ma, a modulating tube current with 10 SD, and a modulating tube current with iterative reconstruction with 5 SD. For craniosynostosis, organ doses were measured for a fixed tube current of 150 ma, a modulating tube current with 10 SD, and a modulating tube current with iterative reconstruction with 10 SD. For trauma follow-up, organ doses were measured for a fixed tube current of 250 ma, a modulating tube current with 5 SD, and a modulating tube current with iterative reconstruction with 5 SD. For VP Shunt, organ doses were measured for a fixed tube current of 200 ma, a modulating tube current with 5 SD, and a modulating tube current with iterative reconstruction with 5 SD. 212

213 9.3.2 Results Measured organ doses are listed in Table 9-7 for various head protocols and in Table 9-8 for the various stealth head protocols. In order to evaluate organ dose reduction by exam indication, Figures 9-14 through 9-17 show the organ doses measured for the standard and acceptable low-dose protocols for hydrocephalus, craniosynostosis, trauma follow-up, and VP shunt, respectively. The average organ dose decrease for hydrocephalus was 42% for the fixed tube current protocol, 66% for the tube current modulation protocol, and 31% for the iterative reconstruction protocol. The average organ dose decrease for craniosynostosis was 42% for the fixed tube current protocol, 66% for the tube current modulation protocol, and 78% for the iterative reconstruction protocol. The average organ dose decrease for trauma follow-up was 9% for the fixed tube current protocol, 18% for the tube current modulation protocol, and 31% for the iterative reconstruction protocol. The average organ dose decrease for VP-shunt was 15% for the fixed tube current protocol, 10% for the tube current modulation protocol, and 16% for the iterative reconstruction protocol. 9.4 Analysis of Pediatric Patients Examined with Head CT at UF Health Methods With approval by the University of Florida Investigational Review Board (Study Number ) a log of 500 patients was generated by searching the radiology picture archiving and communication system (PACS) for any patient under 18 years old who had a head CT exam at UF Health between May 30, 2010 and May 30, The patients medical record numbers (MRNs) were used to locate the CT images correlating to the selected CT study. Once the CT images were located and reviewed using the PACS (Visage Imaging, Inc, San Diego, CA) on radiology department 213

214 computers, all identifying information was removed, with patients assigned a Subject Number from 1 to 300. It was noted whether the patient was male or female, the patient s age during the exam, the indication for the exam, the CT scanner model and exam location at UF Health, and the dose metrics reported by the scanner Results Of the 300 patients who were investigated, 138 were female (46%) and 162 were male (54%). In order to have sufficient data for patients scanned with pediatric and adult head protocols, we selected 150 patients between the ages of newborn to 7 years old, and 150 patients between the ages of 8 years old and 17 years old. Figure 9-18 shows the distribution of indications for head CT exams in pediatric patients. Overall, among the 300 pediatric patients, the most common reasons for scanning the head of a child is trauma and headache, which is in agreement with findings from Miglioretti (14). Out of the 300 patients, 138 (46%) were scanned for trauma, including indications stating trauma, fall, head injury, or motor vehicle collision. Furthermore, 50 (17%) patients were scanned for headache. Other common indications included seizures (24 exams), and altered mental status (17 exams). Exams that were not frequently acquired were grouped into the category other, which included indications such as hemorrhage, cerebral palsy, hypertension, syncope, vomiting, among others. The frequency of scanning children for the low-dose indications we investigated was only 3.7 %. Specifically, 4 patients were scanned for assessing hydrocephalus, 5 patients were scanned for VP-shunt, and 2 patients were scanned for evaluating craniosynostosis. None of the indications stated whether the exam was a follow-up, so it is likely some of the trauma exams were for follow-up. The patient ages, scan protocol, 214

215 scanner model, and exam CTDIvol are summarized in Table 9-9 for the hydrocephalus, VP-shunt, and craniosynostosis exams. The patients examined for hydrocephalus were 4 days old, 1 year old, 1 year old, and 12 years old. The 4-day-old patient presented with concerns for obstructive hydrocephalus, and was scanned with a trauma head CT protocol dedicated for children between 0 to 2 years of age on a Toshiba 64-slice scanner with 120 kvp, fixed tube current of 180 ma, and a CTDIvol of mgy. A 1-year-old patient with a history of hydrocephalus presented with epilepsy and concern for progressive hydrocephalus, and was scanned with a head CT protocol dedicated for children between 0 to 2 years of age on a Toshiba Aquilion ONE, 120 kvp, fixed 200 ma, and a CTDIvol of 34.2 mgy. The other 1-year-old patient presented with sudden widespread attack of inflammation in the brain and was scanned with a stealth head CT scan on a Toshiba Aquilion 16- slice scanner with 120 kvp, fixed 300 ma, and a CTDIvol of 58.2 mgy. The 12-year-old patient had a history of hydrocephalus and presented with an onset of dizziness, nausea, and vomiting. The patient was scanned to evaluate recurring hydrocephalus with a standard head without contrast CT exam on a Toshiba Aquilion PRIME CT scanner with 120 kvp, fixed 270 ma, and a CTDIvol of 70.6 mgy. This 12-year old patient is an example of a child who received an adult head CT exam, but could have been scanned with a lower CTDIvol of 15 mgy while producing acceptable image quality. Although we did not investigate image quality in subjects with smaller heads, we showed that larger sized heads provide acceptable image quality for assessing hydrocephalus with an exam having a CTDIvol of only 15 mgy. Because the other three patients under the age of 8 years old received exams with doses higher than 215

216 15 mgy, and they are younger with smaller heads, it is of great interest to investigate the potential for reducing the dose to these exams in the future. The patients examined for VP shunt evaluation were 5, 7, 11, 13, and 15 years old. The 5-year-old patient had a history of VP-shunt placement and presented with an onset of optic disc swelling caused by increased intracranial pressure. The patient was scanned for evaluation of shunt status with a routine brain CT protocol for children between 3 to 8 years old on a Toshiba 64-slice CT scanner with 120 kvp, 300 ma, and a CTDIvol of 33.5 mgy. The 7-year old patient had a history of VP-shunt placement and presented with hypertension and agitation. The child was scanned for evaluation for shunt malfunction with a stealth head CT protocol on a Toshiba Aquilion PRIME CT scanner with 120 kvp, 270 ma, and a CTDIvol of 32.4 mgy. The 11-year old patient had a history of VP shunt placement presenting with headaches and behavior changes. The patient was scanned with a standard head without contrast CT protocol on a Toshiba 64-slice CT scanner with 120 kvp, 280 ma, and a CTDIvol of 72.2 mgy. The 13-year-old patient presented with a history of VP shunt placement and infection, with a new onset of chronic headaches, and was scanned with a standard head CT without contrast on a Siemens Sensation 16 scanner with 120 kvp, 350 ma, and a CTDIvol of mgy. The 15-year old patient had a history of VP-shunt placement and presented with a headache and pain behind the ears. The patient was scanned for VP shunt malfunction with a stealth head CT protocol on a Toshiba Aquilion PRIME CT scanner with 120 kvp, 270 ma, and a CTDIvol of 32.4 mgy. Although we were told that patients with VP shunts are assessed with stealth head protocols, only two out of the five patients were scanned with a reduced-dose 216

217 stealth head protocol with an exam CTDIvol of 32.4 mgy. The two patients over the age of 8 years old that were scanned with head protocols received exam CTDIvol values of 72.2 mgy and 67.2 mgy. The image quality assessment in the previous section reported that radiologists found VP shunt image quality to be acceptable when using a fixed tube current of 200 mgy, resulting in an exam CTDIvol of 23.7 mgy. As a result, these two patients aged 11 and 13 years old could have benefited from the low-dose protocol that utilizes about one third of the dose, as compared to the exams they received. Of the two craniosynostosis exams, one patient was 7 months old with a history of sagittal craniosynostosis. The patient was scanned on a Toshiba Aquilion 64-slice CT scanner with a dedicated craniosynostosis protocol with 120 kvp and a CTDIvol of 32 mgy. The other patient was 9 years old with a history of craniosynostosis and an onset of chronic increased intracranial pressure. The patient was scanned on a Siemens Sensation 16-slice CT scanner with a head without contrast CT protocol with 120 kvp and a CTDIvol of 67.2 mgy. This pediatric patient is an example of a 9-year old child who could have been scanned with a reduced-dose protocol with an exam CTDIvol as low as 11.8 mgy. 9.5 Discussion With the new ICRP 118 findings that cataract threshold is only 500 mgy, as low as ten Head CT scans or six Trauma Head and Cervical Spine CT scans would meet the threshold for cataract induction (11). Furthermore, children are more radiosensitive to cataract induction than adults (135). Based on the brain dose measurement of 43.5 mgy, it would take less than two head CT examinations to reach the 50 mgy level, which was estimated to put children at 3 times increased risk of brain cancer (118). 217

218 Our measured brain and thyroid doses were higher than those estimated by Miglioretti for children between ages of 10 to 14 years old, shown in Table 9-10 (14). However, without knowledge of the scan parameters collected by Miglioretti, it is difficult to compare measured and calculated organ doses, as it is likely that the head CT examinations conducted across the 5 hospital systems used a variety of CT scanners and scan protocol settings. Surveys done in the UK showed that minimum and maximum doses for head CT examinations for 10-year-old children had an examination CTDIvol ranging from 32 mgy to 51 mgy (136). The exam CTDIvol used on 10-year old children at UF Health is significantly higher, at 70.8 mgy, surpassing the maximum CTDIvol found in the UK. This may explain why our measured organ doses were higher than those calculated by Miglioretti, and also suggests the potential for reducing the scan parameters for 10-year old children at UF Health. The 75 th percentile CTDIvol value for adult head exams in the UK study was 66 mgy, and for pediatric head exams for 10-year-old children was 46 mgy. This highlights that not all institutions utilize adult protocols on 10-year old patients. Our measured organ doses were slightly lower than those calculated with Lee et al s organ dose conversion coefficients for a 10-year old patient, also shown in Table 9-2 (104). However, our measured skin dose was significantly higher at mgy, as compared to Lee s calculated 6.66 mgy. This is due to Lee s skin dose averaged over the entire skin tissue, whereas our skin dose was only measured in the region of irradiated skin. This measurement is more important for monitoring deterministic effects to the skin from cumulative radiation doses. The measured salivary glands dose was 218

219 also higher than Lee s doses, which could be due to a variety of factors including position of the salivary glands in the gantry, as well as effects of the bow tie filter. A limitation of this study is that image quality analysis was based on normal anatomical structures in the head. Because it is difficult to obtain postmortem subjects with pathologies such as craniosynostosis, hydrocephalus, VP shunt, or trauma followup, we were not able to have radiologists evaluate their confidence in the specific pathologies themselves. However, the radiologists stated that the questions asked, such as GWM differentiation, and confidence in assessing the pathologies, provided a good alternative for evaluating heads without pathologies present. Although exam parameters may be modified on a case by case basis, CT management expressed concerns regarding how the low-dose CT protocols would be utilized by the CT technologists. They expressed it would be best to have a global lowdose head CT protocol that could be referred to in such protocols. Since the VP shunt protocol already utilizes a lower dose scan protocol (29.6 mgy) than a standard head protocol (70.8 mgy), our findings revealed only small levels of dose reduction were allowed, and perhaps was not worth modifying for now. Furthermore, because trauma follow-up allowed significantly less dose reduction than hydrocephalus and craniosynostosis, and especially considering the exam indications did not state if a patient is being examined for a follow-up, we decided to remove trauma follow-up from the global low dose protocol. As a result, we recommended using the limiting dose reduction of tube current modulation with 10 SD for both hydrocephalus and craniosynostosis, producing about 66% dose reduction. 219

220 Although the recommendations to use tube current modulation with 10 SD were valid for one postmortem subject, the assessment needs to be investigated with additional heads with varying sizes and attenuation in order to have an understanding of how the tube current will behave. Limitations to this approach include the unfamiliarity of using tube current modulation in patient heads. Setting a safe minimum and maximum tube current level can ensure the patient won t receive unnecessarily high or low levels of tube current. It is recommended to set the limit at the original fixed ma value of 270 ma for head and 250 ma for stealth head, and to agree upon a safe level of minimum ma. Another conservative approach that would not introduce the effects of tube current modulation would be to utilize a standard head protocol with a reduced fixed tube current of 150 ma for hydrocephalus or craniosynostosis protocols, which also produced acceptable image quality and produce about 44% dose reduction. Until additional heads are scanned with tube current modulation, it may be better to use a fixed tube current of 150 ma. Although the initial exam indication states the reason for acquiring the CT exam, it does not necessarily explain the final outcome of the exam. For example, if a patient s indication stated headache, and the radiologist identified hydrocephalus, that exam would have been grouped into the headache indication in Figure Alternatively, if a patient s indication stated evaluate for shunt malfunction, and the radiologist identified a tumor, the exam would have been grouped into the VP shunt indication. The reason for this is the initial indication for scanning is what determines what scan protocol will be utilized on the patient. As a result, only those patients with suspicious hydrocephalus, 220

221 VP shunt, trauma follow-up, or craniosynostosis would benefit from a low-dose protocol, and out of 300 patients, only about 11 patients would have benefited from low-dose protocols. However, regardless of how infrequent these indications arise, it remains crucial to apply the ALARA principle for these patients who can afford dose reduction. Furthermore, it is possible that a higher frequency of patients might be affected with a greater study population. Although 300 patients is a relatively small sample size compared to the number of patients who undergo CT examinations per year, we were limited to the timeconsuming tasks of manually scrolling through each patient s scan series. Future studies should conduct the assessment with a greater sample size. 221

222 A B C Figure 9-1. Images of a PVC tube placed in third ventricle in the (A) axial, (B) coronal, and (C) sagittal planes viewed in a soft tissue window to display the third ventricle. Figure courtesy of author. A B Figure 9-2. Images of a PVC tube and OSL dosimeters placed in the third ventricle in the (A) coronal and (B) sagittal planes viewed in a bone window to display the dosimeter placement in the brain. Figure courtesy of author. 222

223 A B Figure 9-3. Scan ranges for head CT protocols at UF Health. The scan ranges are defined by the Department of Radiology. A) The scan range for the Head CT protocol begins at the skull vertex and ends at the skull base (red lines). B) The scan range for the Trauma Head CT protocol begins at the skull vertex and ends at the chin (red lines), and for the Trauma Head and Cervical Spine protocol follows with a modulating cervical spine protocol beginning at the sella and ending at the T3 vertebra (blue lines). 223

224 A B C Figure 9-4. Screenshots of scan ranges for the head CT protocols acquired on Subject 5. The scan ranges were selected by an experienced CT technologist based on the defined protocols by the Department of Radiology. A gantry tilt of 15- degrees was used to reduce intracranial artifact. A) The scan range is shorter for the Head CT protocol than for the B) Trauma Head CT protocol. C) The sagittal view of the cervical spine is also shown to illustrate the cervical spine imaged down to the third uppermost thoracic vertebra, T3, during the Trauma Head and Cervical Spine CT protocol. Figure courtesy of author. 224

225 Table 9-1. Organ doses (mgy) measured for a standard Head CT protocol and Trauma Head and C-Spine CT protocol in Subject 7 CT Protocol Head Dose (mgy) Salivary Glands Organs Lens Skin Brain Thyroid Avg Max SD Trauma Head Avg Max SD Trauma Head and Cervical Spine Avg Max SD Note: Organ doses are reported as the average (Avg), maximum (Max) and standard deviation (SD) of all organ dose measurements acquired in each organ. 225

226 Organ Dose (mgy) Lens Salivary Glands Skin Brain Thyroid Head Trauma Head Trauma Head and Cervical Spine CT Protocol Figure 9-5. Average organ doses (mgy) measured for typical head CT protocols in Subject 5 226

227 CTDI vol (mgy) Standard Fixed ma TCM AIDR Dose Reduction Method Figure 9-6. CTDIvol values resulting from modifying scan parameters in head protocols. Dose reduction was achieved by lowering the tube current from 270 ma to fixed tube current values (labeled as Fixed ma ) of 250 ma (blue), 200 ma (red), and 150 ma (green), enabling tube current modulation with different target noise levels (labeled as TCM ) of 5 SD (blue), 7.5 SD (red), and 10 SD (green), and enabling iterative reconstruction with different target noise levels (labeled as AIDR ) of 2.5 SD (blue), 5 SD (red), and 7.5 SD (green). 227

228 CTDIvol (mgy) Standard Fixed ma TCM AIDR Dose Reduction Method Figure 9-7. CTDIvol values resulting from modifying scan parameters in Stealth Head CT protocols. Dose reduction was achieved by lowering the tube current from 250 ma to fixed tube current values (labeled as Fixed ma ) of 200 ma (blue), 150 ma (red), and 100 ma (green), enabling tube current modulation with different target noise levels (labeled as TCM ) of 5 SD (blue), 7.5 SD (red), and 10 SD (green), and enabling iterative reconstruction with different target noise levels (labeled as AIDR ) of 5 SD (blue), 7.5 SD (red), and 10 SD (green). 228

229 A B C Figure 9-8. Graph of tube current of Head protocol with increasing target noise levels of A) 5 SD, B) 7.5 SD, and C) 10 SD. Tube current is graphed in red, modulating along the z-axis as the attenuation of the head changes. Figure courtesy of author. A B C Figure 9-9. Graph of tube current of Stealth protocol with increasing target noise levels of A) 5 SD, B) 7.5 SD, and C) 10 SD. Tube current is graphed in red, modulating along the z-axis as the attenuation of the head changes. Figure courtesy of author. 229

230 A B C Figure Graph of tube current of Head protocol with iterative reconstruction with increasing target noise levels of A) 2.5 SD, B) 5 SD, and C) 7.5 SD. Tube current is graphed in green, modulating along the z-axis as the attenuation of the head changes. Figure courtesy of author. A B C Figure Graph of tube current of Stealth CT protocol with iterative reconstruction with increasing target noise levels of A) 5 SD, B) 7.5 SD, and C) 10 SD. Tube current is graphed in green, modulating along the z-axis as the attenuation of the head changes. Figure courtesy of author. 230

231 Figure Graph of tube current of Head protocol with fixed tube current (red), tube current modulation (blue) and iterative reconstruction (orange) with increasing target noise levels. Figure courtesy of author. Figure Graph of tube current of Stealth CT protocol demonstrating the effects of mis-centered patients. Correctly centered scan is shown in green, and offcentered scan by 4 cm anteriorly is shown in red. Figure courtesy of author. 231

232 Table 9-2. Radiologists utilized to assess image quality of pediatric CT scans Radiologist ID Radiologist Specialty Radiologist Name 1 Pediatric neuroradiology Dhanashree Rajderkar 2 Adult neuroradiology Jeffrey Bennett Table 9-3. Image quality scores for diagnosing hydrocephalus with 1 standard and 10 low-dose head CT protocols in Subject 5 Image quality feature CSF over the brain R- ID Fixed ma (ma) TCM (SD) IR (SD) Temporal horns Frontal horns Third ventricle Periventricular interstitial edema Sylvian fissures Confidence in diagnosing hydrocephalus Note: Radiologists are described by their Radiologist ID (R-ID) described in Table 9-2. Fixed ma refers to protocols acquired with a fixed tube current (ma), TCM refers to protocols acquired with modulating tube current driven by a desired standard deviation (SD) of image quality, and IR refers to protocols acquired with modulating tube current and iterative reconstruction (IR) driven by a desired standard deviation (SD) of image quality. The standard Head CT protocol is given by the fixed 270 ma protocol. 232

233 Table 9-4. Image quality scores for diagnosing craniosynostosis with 1 standard and 10 low-dose head CT protocols in Subject 5 Image quality feature Shape of head Sutures Cortical and trabecular bone structures Air filled compartments R- ID Fixed ma (ma) TCM (SD) IR (SD) Confidence in diagnosing craniosynostosis Note: Radiologists are described by their Radiologist ID (R-ID) described in Table 9-2. Fixed ma refers to protocols acquired with a fixed tube current (ma), TCM refers to protocols acquired with modulating tube current driven by a desired standard deviation (SD) of image quality, and IR refers to protocols acquired with modulating tube current and iterative reconstruction (IR) driven by a desired standard deviation (SD) of image quality. The standard Head CT protocol is given by the fixed 270 ma protocol. 233

234 Table 9-5. Image quality scores for diagnosing trauma follow-up with 1 standard and 10 low-dose head CT protocols in Subject 5 Image quality feature GW Matter differentiation W Matter abnormalities Extra-axial fluid space Ventricular size R-ID Fixed ma (ma) TCM (SD) IR (SD) Confidence in diagnosing trauma follow up Note: Radiologists are described by their Radiologist ID (R-ID) described in Table 9-2. Fixed ma refers to protocols acquired with a fixed tube current (ma), TCM refers to protocols acquired with modulating tube current driven by a desired standard deviation (SD) of image quality, and IR refers to protocols acquired with modulating tube current and iterative reconstruction (IR) driven by a desired standard deviation (SD) of image quality. The standard Head CT protocol is given by the fixed 270 ma protocol. 234

235 Table 9-6. Image quality scores for diagnosing VP-shunt with 1 standard and 9 lowdose head CT protocols in Subject 5 Image quality feature Temporal horns Sharpness of ventricle outline R-ID Fixed ma (ma) TCM (SD) IR (SD) Confidence in diagnosing VP shunt Note: Radiologists are described by their Radiologist ID (R-ID) described in Table 9-2. Fixed ma refers to protocols acquired with a fixed tube current (ma), TCM refers to protocols acquired with modulating tube current driven by a desired standard deviation (SD) of image quality, and IR refers to protocols acquired with modulating tube current and iterative reconstruction (IR) driven by a desired standard deviation (SD) of image quality. The standard Head CT protocol is given by the fixed 250 ma protocol. 235

236 Table 9-7. Organ doses measured for standard and low-dose Head CT protocols in Subject 5 CT Protocol Fixed ma 270 ma Organs Dose (mgy) Salivary Brain Thyroid Lens Skin Glands Avg Max SD Fixed ma 250 ma Fixed ma 150 ma TCM 5 SD TCM 10 SD IR 5 SD Avg Max SD Avg Max SD Avg Max SD Avg Max SD Avg Max SD IR 10 SD Avg Max SD Note: Organ doses are reported as the average (Avg), maximum (Max) and standard deviation (SD) of all organ dose measurements acquired in each organ. Fixed ma refers to protocols acquired with a fixed tube current (ma), TCM refers to protocols acquired with modulating tube current driven by a desired standard deviation (SD) of image quality, and IR refers to protocols acquired with modulating tube current and iterative reconstruction (IR) driven by a desired standard deviation (SD) of image quality. The standard Head CT protocol is given by the fixed 270 ma protocol. 236

237 Table 9-8. Organ doses measured for standard and low-dose Stealth Head CT protocols in Subject 5 CT Protocol Fixed ma 250 ma Organs Dose (mgy) Salivary Brain Thyroid Lens Skin Glands Avg Max SD Fixed ma 200 ma TCM 5 SD Avg Max SD Avg Max SD IR 5 SD Avg Max SD Note: Organ doses are reported as the average (Avg), maximum (Max) and standard deviation (SD) of all organ dose measurements acquired in each organ. Fixed ma refers to protocols acquired with a fixed tube current (ma), TCM refers to protocols acquired with modulating tube current driven by a desired standard deviation (SD) of image quality, and IR refers to protocols acquired with modulating tube current and iterative reconstruction (IR) driven by a desired standard deviation (SD) of image quality. The standard Head CT protocol is given by the fixed 250 ma protocol. 237

238 Organ Dose (mgy) Organ Dose (mgy) 70 Lens Skin Salivary Glands Brain Bone Thyroid Standard (270 ma) LD (150 ma) TCM (10 SD) AIDR (5 SD) CT Protocol Figure Average organ doses measured from standard and acceptable reduceddose protocols for diagnosing hydrocephalus. LD refers to a fixed ma, TCM refers to tube current modulation, and AIDR refers to iterative reconstruction. 70 Lens Salivary Glands Skin Brain Bone Thyroid Standard (270 ma) LD (150 ma) TCM (10 SD) AIDR (10 SD) CT Protocol Figure Average organ doses measured from standard and acceptable reduceddose protocols for diagnosing craniosynostosis. LD refers to a fixed ma, TCM refers to tube current modulation, and AIDR refers to iterative reconstruction. 238

239 Organ Dose (mgy) Dose (mgy) Lens Skin Salivary Glands Brain Bone Thyroid Standard (270 ma) LD (250 ma) TCM (5 SD) AIDR (5 SD) CT Protocol Figure Average organ doses measured from standard and acceptable reduceddose protocols for diagnosing trauma follow-up. LD refers to a fixed ma, TCM refers to tube current modulation, and AIDR refers to iterative reconstruction. 25 Salivary Glands Lens Skin Brain Bone Thyroid Standard 250 ma Low Dose 200 ma TCM 5 SD AIDR 5 SD CT Protocol Figure Average organ doses measured from standard and acceptable reduceddose protocols for diagnosing VP Shunt. LD refers to a fixed ma, TCM refers to tube current modulation, and AIDR refers to iterative reconstruction. 239

240 Hydorcephalus 1% Craniosynostosis 1% VP Shunt 2% Apnea 2% Hemorrhage 3% Altered Mental Status 6% Trauma 46% Seizure 8% Other 14% Headache 17% Figure Distribution of exam indications for pediatric head CT at UF health Table 9-9. Clinical exam parameters for pediatric patients scanned with CT for craniosynostosis, hydrocephalus, and VP Shunt at UF Health Exam Indication Age CT Protocol Name CT Scanner CTDIvol 7 m Craniosynostosis Toshiba Aquilion 32.0 Craniosynostosis 9 y Head w/o Siemens Sensation Hydrocephalus VP Shunt 4 d 0-2 Trauma Head Toshiba Aquilion y Stealth Toshiba Aquilion y Brain Baby 0-2yrs Toshiba Aquilion ONE y Head w/o Toshiba Aquilion PRIME y Brain Child 3-8 Toshiba Aquilion y Stealth Toshiba Aquilion PRIME y Head w/o Toshiba Aquilion y Head w/o Siemens Sensation y Stealth Toshiba Aquilion PRIME