In routine clinical practice, the majority of single-source abdominal computed tomographic (CT) examinations are performed with a tube potential of 12

Transcription

1 Note: This copy is for your personal, non-commercial use only. To order presentation-ready copies for distribution to your colleagues or clients, contact us at ORIGINAL RESEARCH n MEDICAL PHYSICS Appropriate Patient Selection at Abdominal Dual-Energy CT Using 80 kv: Relationship between Patient Size, Image Noise, and Image Quality 1 Luís S. Guimarães, MD2 Joel G. Fletcher, MD William S. Harmsen, MS Lifeng Yu, PhD Hassan Siddiki, MD Zachary Melton, BA James E. Huprich, MD David Hough, MD Robert Hartman, MD Cynthia H. McCollough, PhD 1 From the Department of Radiology, Mayo Clinic, 200 First St SW, Rochester, MN Received October 29, 2009; revision requested December 18; revision received April 19, 2010; accepted May 20; fi nal version accepted July 7. Address correspondence to J.G.F. ( fl etcher.joel@mayo.edu ). 2 Current address: Department of Radiology, Hospital S. Teotónio, Viseu, Portugal. Purpose: Materials and Methods: Results: Conclusion: To determine the computed tomographic (CT) detector configuration, patient size, and image noise limitations that will result in acceptable image quality of 80-kV images obtained at abdominal dual-energy CT. The Institutional Review Board approved this HIPAAcompliant retrospective study from archival material from patients consenting to the use of medical records for research purposes. A retrospective review of contrast material enhanced abdominal dual-energy CT scans in 116 consecutive patients was performed. Three gastrointestinal radiologists noted detector configuration and graded image quality and artifacts at specified levels midliver, midpancreas, midkidneys, and terminal ileum by using two five-point scales. In addition, an organ-specific enhancementto-noise ratio and background noise were measured in each patient. Patient size was measured by using the longest linear dimension at the level of interest, weight, lean body weight, body mass index, and body surface area. Detector configuration, patient sizes, and image noise levels that resulted in unacceptable image quality and artifact rankings (score of 4 or higher) were determined by using multivariate logistic regression. A mm detector configuration resulted in fewer images with unacceptable quality than did the mm configuration at all anatomic levels ( P =.004,.01, and.02 for liver, pancreas, and kidneys, respectively). Image acceptability for the kidneys and ileum was significantly greater than that for the liver for all readers and detector configurations ( P,.001). For the mm detector configuration, patient longest linear dimensions yielding acceptable image quality across readers ranged from 34.9 to 35.8 cm at the four anatomic levels. An 80-kV abdominal CT can be performed with appropriate diagnostic quality in a substantial percentage of the population, but it is not recommended beyond the described patient size for each anatomic level. The mm detector configuration should be preferred. q RSNA, 2010 q RSNA, radiology.rsna.org n Radiology: Volume 257: Number 3 December 2010

2 In routine clinical practice, the majority of single-source abdominal computed tomographic (CT) examinations are performed with a tube potential of 120 kv. However, several studies have shown that it is possible to reduce the tube energy to kv and still have acceptable image quality without losing low-contrast detectability ( 1 3 ). The most substantial advantages of reducing the tube energy in single-source CT are potential radiation dose reduction ( 4 7 ) and increase in image contrast, which is more pronounced for materials with high effective atomic number, such as iodine and bone ( 2,4,7 9 ). Improved image enhancement for iodine increases the conspicuity of enhancing lesions and structures, even allowing the image contrast to be preserved when the contrast media load or the rate of injection is reduced ( 9,10 ). A tube potential of 80 kv is also routinely used in dual-energy CT applications ( ). Dual-energy CT utilizes CT data from two different energy spectra to discriminate, display, and characterize tissue composition ( ). In dualenergy CT applications, data sets at low (usually 80 kv) and high (140 kv) energy are acquired simultaneously with alternative processing techniques then applied Advances in Knowledge n Patient size cutoffs can be used to select patients who can successfully undergo 80-kV imaging with acceptable image quality, but cutoff values should be modified according to the organ of interest and the desired scanning protocol. n Increased image noise level caused by lower energy scanning results in less subjective image quality compromise when interpreting the kidney and ileum than when interpreting the liver and pancreas. n The cm detector configuration at abdominal dualsource, dual-energy CT results in significantly better image quality than the mm detector configuration. to obtain material-specific information ( 12,13,15,16 ). Additionally, by combining data from low- and high-energy data sets, a single set of blended-voltage images can be generated to be used for routine diagnostic purposes. These mixed-voltage images appear similar to conventional single-energy scans, but demonstrate increased contrast due to the contribution of the 80-kV data set, which often allows for improved visualization of vascular structures and diseases demonstrated by differential contrast enhancement compared with the surrounding tissues ( 12,16 ). The most notable disadvantage of low-voltage CT is the potential for increased image noise ( 4,5,7,9,17 ) resulting from the lower tube energy, which results in unacceptable images in large patients. Beyond a certain patient size, the benefit of increased contrast and increased contrast-to-noise ratio (CNR) at 80 kv is offset by the increased noise and beam-hardening artifacts, making abdominal 80-kV single-source and dual-energy CT (using 80 kv in the lower energy tube) inadvisable. Our purpose was to determine the CT detector configuration, patient size, and image noise limitations that will result in acceptable image quality of 80-kV images obtained at abdominal dual-energy CT. This knowledge would permit an improved selection of patients for both dual-energy and single-source low-voltage imaging and facilitate successful investigation of new clinical applications. Materials and Methods The Institutional Review Board approved this Health Insurance Portability and Implication for Patient Care n An 80-kV abdominal CT scan, which allows for radiation dose reduction, increased image contrast, and dual-energy imaging, can be performed with acceptable image quality when careful attention is paid to acquisition techniques, patient size, and organ of interest. Accountability Act-compliant retrospective study conducted from data found in institutional patient databases and archives. Signed consent from the patient to use past medical data for research purposes was obtained. Patients Between March 2007 and March 2008, 153 consecutive patients who consented to the use of past medical records for research purposes underwent an abdominal contrast material enhanced dualenergy CT. Inclusion criteria included a section thickness of 2 3 mm and archived 80-kV images that were available for review. Exclusion criteria included image unacceptability due to motion or causes unrelated to the acquisition technique or patient size. Thirtyfive patients were excluded because they did not have the 80-kV images archived. One patient was excluded because of artifacts caused by the arms overlying the abdomen, and another one was excluded because of severe artifacts caused by beam-hardening effects from excessively dense oral contrast material. Therefore, our study cohort consisted of 116 patients (mean age, 59.7 years; range, years), which included 78 men (mean age, 61.7 years; range, years) and 38 women (mean age, 55.6 years; range, Published online before print /radiol Radiology 2010; 257: Abbreviations: BMI = body mass index CNR = contrast-to-noise ratio CTDI vol = volume CT dose index ROI = region of interest Author contributions Guarantors of integrity of entire study, L.S.G., J.G.F.; study concepts/study design or data acquisition or data analysis/ interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of fi nal version of submitted manuscript, all authors; literature research, L.S.G., J.G.F., L.Y., H.S., C.H.M.; clinical studies, L.S.G., J.G.F., H.S., J.E.H., D.H., R.H.; experimental studies, L.S.G., J.G.F., H.S., C.H.M.; statistical analysis, J.G.F., W.S.H., H.S.; and manuscript editing, L.S.G., J.G.F., W.S.H., L.Y., J.E.H., D.H., R.H., C.H.M. Authors stated no fi nancial relationship to disclose. Radiology: Volume 257: Number 3 December 2010 n radiology.rsna.org 733

3 20 86 years). The most frequent indications for the examinations were hematuria (25.9%, 30 of 116), kidney stone disease (14.7%, 17 of 116), suspicion or re-evaluation of hepatocellular carcinoma (13.8%, 16 of 116), renal masses (12.9%, 15 of 116), pancreatic mass (5.2%, six of 116), or Crohn disease (5.2%, six of 116). Patient size was estimated by using several methods. Each patient s weight and body mass index (BMI) were extracted from review of the medical records. Lean body weight and body surface area were calculated as previously described ( 18,19 ). Patient size was also estimated by using the longest linear axial dimension (skin to skin), which was measured in each of the selected four sections. We chose the longest linear axial dimension to represent patient size, because this length primarily determines the attenuation level of the patient and, if reaching a certain level, results in streaking artifacts and photon starvation ( 20 ). CT Examinations The selected CT examinations were performed with a dual-source CT scanner (Somatom Definition; Siemens Medical Systems, Forchheim, Germany) operating in dual-energy mode, with one tube using 80 kv and the other using 140 kv. Automatic-exposure control software (CareDose4D; Siemens Medical Systems) was turned on for both tubes, with quality reference milliampere-seconds that ranged from 80 to 200 mas for the 140-kV tube and from 340 to 611 mas for the 80-kV tube, depending on the type of examination performed. The quality reference milliampere-seconds were set up in such a way that the total volume CT dose index (CTDI vol ) of the dual-energy scan was equivalent to that of single-energy abdominal examinations. With the aforementioned default quality reference milliampere-seconds settings, the CTDI vol of 80 kv was about 42% 50% of the total CTDI vol. Because of the use of automatic exposure control, the tube currents were modulated for different patient sizes independently and thus the actual total CTDI vol and the distribution between the two tubes varies slightly for different patient sizes. In larger patients, the maximum tube current for the 80-kV tube (500 ma) is reached more often than the maximum tube current for the 140-kV tube (500 ma) because of the higher tube current time product setting in the 80-kV tube. Gantry rotation time was 0.5 second for all patients. The pitch was 0.5 for all examinations except for CT urography, which was performed with a pitch of 0.7. All patients received 140 ml of iohexol (Omnipaque 300; GE Health Care, Milwaukee, Wis) injected intravenously at a rate of 4 5 ml/sec. The phase of enhancement (arterial, enteric, portal, or delayed) for the dualenergy portion of each examination was recorded. At our institution, we select the phase of enhancement most likely to produce the greatest enhancement differences between the suspected disease and the background organ for dualenergy acquisition. Phases of enhancement therefore selected for this study included the late arterial phase for liver examinations, the enteric phase for CT enterography, the pancreatic phase for pancreatic examinations, and the portal (70 seconds) or delayed phase (90 seconds) for CT urography. Two-detector configurations were used: mm and mm. The mm detector configuration has a physical size of mm. With a z-flying focal spot, sampling along z direction is doubled. There is no z-flying focal spot with mm collimation. Because of the largerdetector bin size, the mm configuration is theoretically less sensitive to electronic noise. Additionally, with the mm configuration there are several detector bins on each side of the detector measuring the scatter, which allows the utilization of an online cross-scattering correction technique (more efficient than the algorithm used with the mm configuration). Patients were scanned by using one or the other detector configuration, according to the preference of the reading radiologist. The images were then analyzed on a workstation (Leonardo; Siemens Medical Systems, Forchheim, Germany). Qualitative and Quantitative Analyses A qualitative analysis of image quality was performed by using images obtained at four different anatomic levels (midliver, midpancreatic body, midkidneys, and terminal ileum). These images were selected by a gastrointestinal radiologist (J.G.F., with more than 10 years of experience in abdominal CT). In addition, a quantitative analysis of the ratio of the degree of enhancement to image noise was calculated for each of the four selected organs on these images. The qualitative analysis was independently and separately performed by three abdominal radiologists (J.G.F., D.H., and R.H., with 10, 12, and 8 years of experience, respectively), who were blinded to detector configuration. They were asked to grade the overall image quality and image artifacts and reader confidence by using two five-point scales. These two scales were used because it was believed that image quality might be compromised due to factors other than image noise and artifacts. For image quality, the scale was as follows: score of 1, excellent quality (similar to image quality from 16-section multidetector row CT); score of 2, good quality but perceptible differences compared with a 16-section multidetector CT); score of 3, fair but compromised quality; score of 4, poor quality; and score of 5, severe distortion. For image artifacts and reader confidence, the subsequent scale was used as follows: score of 1 = no artifacts, high confidence in diagnostic capability; score of 2 = mild artifacts, no change in confidence; score of 3 = moderate artifacts, decreased confidence but diagnosis still possible; score of 4 = severe artifacts, confidence degraded, diagnosis questionable; and score of 5 = severe artifacts, nondiagnostic. The image quality and image artifacts and reader confidence were considered unacceptable when either the image quality or the artifact and confidence score was greater than a score of 3 by any of the three readers. When all of the readers graded image quality or image artifacts and reader confidence with a score of 1 3, the image was considered acceptable. Figures 1 and 2 are examples of images that were assigned by 734 radiology.rsna.org n Radiology: Volume 257: Number 3 December 2010

4 reader 1 a score of 1 and 1 and 5 and 5 for image quality and artifacts and reader confidence, respectively. The quantitative analysis was performed at a commercially available workstation (Advantage for Windows, version 4.2; GE Health Care) by a gastrointestinal radiologist (J.G.F) by drawing regions of interest (ROIs) with at least 150 pixels in each of the four selected organs and measuring the mean attenuation value (in Hounsfield units) to reflect the organ enhancement and the standard deviation of the mean attenuation value to reflect the organ noise, as both organ enhancement and noise affect lesion conspicuity and image interpretation. We strove to place the ROIs in areas of the organs as homogeneous as possible, avoiding vessels. In the liver, the ROI was placed in the hepatic parenchyma at the level of the left portal vein (avoiding the vessels), and in the kidneys it was drawn in the renal parenchyma (or the cortex only, if the image had been acquired in the arterial phase). At the level of the pancreas, because of its intrinsic heterogeneity and the interdigitation of retroperitoneal fat with the pancreatic parenchyma, organ enhancement was measured in the pancreatic neck or body, while organ noise was measured in adjacent retroperitoneal fat. At the level of the terminal ileum, for similar reasons, organ enhancement was measured in the bowel wall, whereas organ noise was evaluated in the adjacent intraluminal fluid. In patients whose terminal ileum was not filled with fluid, an image with another ileal loop filled with fluid was selected to measure organ noise. If there were no loops with intraluminal fluid or if positive oral contrast material had been used, a collapsed loop was selected and the enhancement and noise values were obtained from a single ROI localized in the collapsed wall. The values of organ enhancement and organ noise were used to calculate organ-specific enhancementto-noise ratios, defined as mean attenuation of the ROI drawn in the specific organ divided by the standard deviation of the mean attenuation (or noise) value of the same ROI (in the liver and kidneys) or the ROI drawn in the retro- Figure 1 Figure 1: (a d) Eighty-kilovolt images obtained as part of dual-energy CT enterographic examination in a 23-year-old man by using mm detector confi guration and 3-mm section thickness at the level of the liver, pancreas, kidneys, and terminal ileum, respectively. Image quality and artifact ranking received a score of 1 at each level by using a predefi ned subjective scale. The longest transverse linear dimension was 29 cm at the level of the liver. peritoneal fat (for the pancreas) and intraluminal fluid (for the ileum). Additionally, a global assessment of image noise was taken by measuring the noise in the subcutaneous abdominal fat for each patient (average of three ROI measurements). Statistical Analysis To determine if one-detector configuration provides better image quality, a single binary outcome variable (image quality and artifacts rated as acceptable versus unacceptable) was created by using image quality and artifact scores given by each radiologist. A x 2 or Fisher exact test (as appropriate) was used to compare the proportions of acceptable images within the mm and mm detector configuration cohorts. The rationales for combining these two scores were as follows: (a) It was believed image acceptability could not be summarized by an assessment of artifacts and confidence alone, (b) there may be some factors that degrade image quality that are not perceived as artifacts (eg, cross-scatter, higher iodine contrast, and noise), and (c) image quality does not reflect diagnostic confidence. This binary outcome variable reflected image acceptability if image quality and image artifacts and reader confidence scores were lower than a score of 4 or conversely image unacceptability if one or both scores were 4 or 5. To compare image acceptability among the four anatomic levels in the same patient, an analysis was performed by using Friedman test to account for withinpatient correlations, and pair-wise comparisons were performed by using Wilcoxon signed rank test. This approach can be thought of as 116 blocks (subjects), Radiology: Volume 257: Number 3 December 2010 n radiology.rsna.org 735

5 Figure 2 Figure 2: (a d) Eighty-kilovolt images obtained at dual-energy CT urography in a 59-year-old man by using mm detector confi guration and 3-mm section thickness at the level of the liver, pancreas, kidneys, and terminal ileum, respectively. The longest transverse linear dimension was 40 cm at the level of the liver. Image quality and artifact ranking received a score of 5 for the liver, pancreas, and terminal ileum and a score of 4 for the kidneys by using a predefi ned subjective scale. sion model, including independent predictors of patient size (as reflected by weight, lean body weight, BMI, body surface area, longest linear dimension), adjusting for the quality reference milliampere-seconds. Logistic regression models were also created by using binary acceptability as the dependent variable and enhancement-to-noise ratio and image noise (in the subcutaneous fat noise) as the independent variable, also adjust ing for the quality reference milliampereseconds. Logistic regression results are reported as odds ratios for image unacceptability, along with 95% confidence intervals and P values. By utilizing the predicted probabilities from these models, the area under the receiver operator characteristic curve was estimated. The association of detector configuration, comparing mm configuration relative to the mm configuration, and image acceptability was also assessed with logistic regression analysis, and results were reported as odds ratios for the unacceptable examination. We report the cutoff values for patient size and image noise above which image quality becomes unacceptable, such that the sensitivity of correctly predicting an unacceptable image is at least 90%. Statistical analysis was performed with software (SAS, version 9.2; SAS Institute, Cary, NC), and a P value equal to or less than.05 was considered to indicate a significant difference. cating slight agree ment, indicating fair agreement, indicating moderate agreement, indicating substantial agreement, and indicating almost perfect agreement ( 21 ). Because we sought to create robust guidelines for patient size criteria at lowvoltage imaging that would be acceptable to many radiologists, we subsequently rated every case as unacceptable if any of the three radiologists rated it as unacceptable. This binary outcome variable (rating an image as unacceptable when it was rated unacceptable by any of the readers) was used as the dependent variable in the logistic regreswith four scores in each block (ie, the acceptability assessment on a scale of 1 5, one value for each of the four anatomic levels). In the assessment of image acceptability, we separately considered detector configuration, reader, and organ of interest, with the Friedman test taking into account the acceptability decision at the four levels in each patient. The concordance between the image quality score and the artifact and confidence score is also reported in terms of their prediction of image acceptability. A multireader concordance correlation coefficient is reported for the agreement of image quality and artifact scores for each organ, with indi- Results Seventy-four of the patients were scanned with the mm detector configuration and 42 were scanned with the mm detector configuration. A total of 97% (113 of 116) of the individuals had a longest linear axial dimension that was coincident with the lateral width. The median longest linear axial dimensions for the mm detector configuration ranged from 38.6 to 40 cm, depending on the anatomic level. For the mm configuration, the correspondent median longest linear axial dimensions ranged from 33 to 35.4 cm, being significantly smaller ( P,.05) at the level of the liver, pancreas, and 736 radiology.rsna.org n Radiology: Volume 257: Number 3 December 2010

6 In the qualitative analysis of image acceptability (ie, the dichotomized acceptable versus unacceptable result based on image quality and artifact scores), there were significant differences between the four anatomic levels for both detector configurations and all three readers ( Fig 3 ). For the mm detector configuration, all six pair-wise comparisons of image acceptability between organs were significant for each reader ( P,.001 to P =.05), except for reader 1, who did not find a significant difference in image acceptability between the kidneys and the terminal ileum. For the mm configuration, all three readers found image acceptability for the kidneys and ileum to be better than that for the liver ( P,.001 for all comparisons). Two of the three readers also found image acceptability for the kidneys and terminal ileum to be higher than that for the pancreas ( P,.05). The multireader concordance coefficients for image quality and arti fact scores showed moderate agreement (concordance correlation coefficient: for liver, for pancreas, for kidneys), except for the terminal ileum, which demonstrated fair agreement (concordance correlation coefficient: 0.341). When an image was rated as unacceptable if any reader thought the image was unacceptable, the kidneys and the terminal ileum had fewer than half the number of unacceptable images compared with the liver at both detector configurations ( Table 1 ). When image quality and image artifact scores were compared (116 patients 3 four anatomic levels 3 three readers = 1392 assesskidneys. The range of patient sizes varied markedly (weight, kg; lean body weight, kg; BMI, kg/m 2 ; body surface area, m 2 ; longest linear dimension, cm). For both the mm and the mm detector configurations, more than half of the patients were scanned in the portal phase (57% [24 of 42] and 62% [46 of 74], respectively). Similar patterns were observed for the arterial (29% [12 of 42] and 15% [11 of 74], respectively), enteric (12% [five of 42] and 20% [15 of 74], respectively), and delayed phases (2% [one of 42] and 3% (two of 74], respectively). There was no significant difference between the number of examinations performed in the different phases of contrast enhancement between the detector configuration ( P =.28). There were significant differences in image acceptability between the detector configurations. Readers 1, 2, and 3 scored 27% (79 of 296), 22% (64 of 296), and 9% (26 of 296) of mm detector images as unacceptable, compared with 11% (18 of 168), 3% (five of 168), and 0.6% (one of 168) of the mm detector images, respectively ( P,.001 for all readers). When any of the three radiologists rated the image as unacceptable (in terms of either image quality or artifacts and reader confidence), the qualitative analysis demonstrated that the mm detector configuration rendered significantly fewer unacceptable images than did the mm detector configuration at all anatomic levels except for the terminal ileum, taking the quality reference milliampere-seconds into account ( Table 1 ). Even at the level of the terminal ileum, the odds ratio for unacceptability for the mm compared with the mm detector configuration was 3.71, but the 95% confidence intervals included 1.0 (95% confidence interval: 0.96, 14.39; P =.06). In contrast, the quantitative analysis showed no statistically significant association between the detector configuration and either organ enhancement, enhancement-to-noise ratio, or organ noise at any of the four anatomic levels ( P =.08.98). Table 1 Percentages of Patients with Unacceptable Image Quality When Quality Reference Milliampere-seconds Is Taken into Account Anatomic Level mm Detector * mm Detector * Odds Ratio P Value Liver 21 (9/42) 51 (38/74) 4.00 (1.56, 10.13).004 Pancreas 14 (6/42) 39 (29/74) 3.82 (1.34, 10.83).012 Kidneys 5 (2/42) 20 (15/74) 6.25 (1.28, 30.60).024 Terminal ileum 7 (3/42) 18 (13/74) 3.71 (0.96, 14.39).058 * Data are percentages and data in parentheses are numbers used to calculate the percentages. Odds of a patient scanned with 64 x 0.6-mm detector, relative to the 14 x 1.2-mm detector confi guration, having unacceptable image quality. Numbers in parentheses are 95% confi dence intervals. ments), there was a high level of concordance between the image quality and the image artifact and confidence scores, with these scores being either identical (975 of 1392, 70%) or differing by only one point without a difference in image acceptability versus nonacceptability (371 of 1392, 27%). There were 55 ( n = 1392, 4%) assessments in which a difference in image quality and image artifact scores changed the assessment of image acceptability compared when only one of the scores was used. Image acceptability was not related to the phase of enhancement for either detector configuration for any organ of interest (eight comparisons, P =.09.90). For the mm detector configuration, all measures of patient size (weight, lean body weight, BMI, body surface area, longest linear dimension), as well as the enhancement-to-noise ratio, and the image noise values in the subcutaneous fat were significantly associated with unacceptability when the quality reference milliampere-seconds was adjusted for every organ ( P,.01; Table 2 ), except for BMI as a predictor of kidney image acceptability ( P =.23). For the mm detector configuration, in which there were fewer unacceptable images, there were differences in the association between these measures and unacceptability. All measures of patient size (except for lean body weight) were significantly associated with liver image acceptability ( P.02). Longest linear dimension was the only size measure to also be significantly associated with pancreatic image acceptability (odds Radiology: Volume 257: Number 3 December 2010 n radiology.rsna.org 737

7 Figure 3 Figure 3: Graph shows percentages of acceptable images for each anatomic level with both detector confi gurations. There are signifi cant differences in acceptability between anatomic levels, with the liver and pancreas yielding an inferior number of acceptable images than the kidneys and ileum at both detector confi gurations ( P =.001 for both detector confi gurations and all three readers for comparisons of liver image acceptability versus kidney or ileum; P = for both detector confi gurations and two of three readers for pancreas image acceptability versus kidney or ileum). ratio = 2.00, P =.0309). Noise in the subcutaneous fat was trended toward an association with image acceptability in the pancreas ( P =.0503) and the terminal ileum ( P =.0630), but did not reach statistical significance. Because of the low number of unacceptable renal and ileal images at this detector configuration (5% [two of 42] and 7% [three of 42], respectively; Table 1 ), no measures of size or noise were associated with unacceptability at these locations. Figure 4 shows the relationship between enhancement-to-noise ratio and image noise versus longest linear dimension for both detector configurations at the level of the liver. Table 3 shows the patient size cutoffs using the longest linear dimension above which one would achieve 90% or greater sensitivity for the prediction of an unacceptable image quality at 80 kv (with the dose levels used), meaning our ability to predict an unacceptable image quality. Table 3 also reports the area under the receiver operator characteristic curve that would be associated with the designated patient size cutoff for predicting image acceptability. At the level of the liver, 72% of our cohort (84 of 116) had the longest linear dimension below 35.8 cm. Moreover, 76% (32 of 42) of all examinations performed with the mm detector configuration and 66% (49 of 74) of the examinations performed with the mm detector configuration had acceptable image quality at all anatomic levels. The noise levels that result in unacceptable image quality and artifact scores for both detector configurations are given in Table 4. Patient weight, BMI, and body surface area were associated with unacceptable image quality when both detector configurations were used at the level of the liver only ( Table 5 ). Discussion Our study demonstrates that while acceptable image quality can be obtained at 80 kv within the constraints of dosematched abdominal dual-energy ( kv) CT scanning, there is a substantial risk for unacceptable image quality unless radiologists consider the acquisition technique, patient size, and organ of interest. On the other hand, if these constraints are taken into account, dual-energy CT can be used with increased confidence, since now there are available guidelines for predicting acceptable image quality at 80 kv. Our study demonstrated that the mm detector configuration produces images with significantly better quality than the mm configuration in the liver and pancreas and should be the preferred detector configuration for routine abdominal dual-energy CT for all but the smallest patients. For within-patient correlations, there were significant differences between the image quality at the four anatomic levels, with the quality being superior at the level of the kidneys and terminal ileum than at the level of liver and pancreas. We surmise that the increased noise levels resulting from 80-kV imaging cause less compromise in subjective image quality in these organs because of their high contrast and relative homogeneity compared with those of the liver and pancreas. Indeed, the cutoff values for the kidney and terminal ileum levels for the mm detector configuration are probably underestimated due to the small number of unacceptable images at these anatomic levels. The size cutoffs provided for the different organs and anatomic levels may be used by the radiologist to set protocols for CT directed to any of these specific organs and also for abdominopelvic CT examinations that interrogate several organs. From a practical perspective, however, the size cutoffs for the various organs are within only 1 cm ( cm) when the mm detector configuration is used. Consequently, since the results of this study became available, we have been using 36 cm as our size cutoff for single or multiphase kV dual-energy abdominopelvic CT with mm detector configuration. Patients larger than 36 cm undergo either dual-energy scanning with 738 radiology.rsna.org n Radiology: Volume 257: Number 3 December 2010

8 Table 2 Association between Measures of Patient Size, Organ Enhancement-to-Noise Ratio, and Image Noise and Unacceptability of 80-kV Images When Quality Reference Milliampere-seconds Is Taken into Account mm Detector mm Detector Measure of Patient Size or Image Noise according to Anatomic Level Odds Ratio for Unacceptability * P Value Odds Ratio for Unacceptability * P Value Longest linear dimension Liver (1.345, 4.662) (1.345, 4.662).0038 Pancreas (1.622, 4.508) (1.066, 3.766).0309 Kidneys (1.417, 5.291) (0.759, 7.278).1386 Terminal ileum (1.221, 2.179) (0.874, 3.891).1080 Enhancement-to-noise ratio Liver (0.098, 0.483) (0.393, 1.475).4195 Pancreas (0.234, 0.633) (0.189, 0.963).0402 Kidneys (0.102, 0.592) (0.300, 1.474).3148 Terminal ileum (0.251, 0.714) (0.405, 1.529).4796 Weight Liver (1.067, 1.219) (1.042, 1.246).0043 Pancreas (1.084, 1.275), (0.989, 1.132).1015 Kidneys (1.068, 2.046) (0.944, 1.410).1622 Terminal ileum (1.071, 1.284) (0.966, 1.276).1414 Lean body weight Liver (1.080, 1.269) (0.982, 1.237).0973 Pancreas (1.090, 1.302) (0.986, 1.381).0718 Kidneys (1.070, 1.327) (0.848, 1.423).4749 Terminal ileum (1.059, 1.283) (0.909, 1.379).2878 BMI Liver (1.204, 1.885) (1.125, 2.000).0058 Pancreas (1.242, 2.027) (0.931, 1.354).2273 Kidneys (0.053, ) (0.001, ).5691 Terminal ileum (1.281, 2.411) (0.978, 2.110).0651 Body surface area (per 1/10 unit) Liver (1.532, 3.625), (1.286, 4.933).0071 Pancreas (1.690, 4.904), (0.934, 2.935).0843 Kidneys (1.871, ) (0.599, 6.138).2733 Terminal ileum (1.586, 5.527) (0.728, 4.496).2018 Image noise in subcutaneous fat Liver (1.132, 1.536) (1.027, 1.415).0221 Pancreas (1.149, 1.636) (1.000, 1.437).0503 Kidneys (1.137, 1.638) (0.723, 5.254).1874 Terminal ileum (1.132, 1.590) (0.985, 1.804).0630 * Data in parentheses are 95% confi dence intervals. 100-kV tube energy for the lower energy tube or a single-energy examination at either 100 or 120 kv. It has been previously recognized that the reduction of tube voltage settings in single-energy CT ( 1 3,7,9,10 ) and utilization of dual-energy applications ( 14 ) are largely constrained by patient size and image noise. Given that a patient s longest linear dimension is almost always coincident with the patient s lateral width, which is an easily accessible and measurable parameter before every CT examination, the results of our study allow for a better patient selection for abdominal dual-energy CT, as well as for single-source 80-kV CT. Lateral width was the only measure of patient size that correlated with image acceptability for both the liver and pancreas with use of mm detector configuration. Our results substantiate prior pilot work focusing on the feasibility of the reduction of tube voltage settings in abdominal CT, but to our knowledge, no study has investigated the cutoff values for patient size and image noise beyond which image quality rendered by low-tube-voltage images is clinically unacceptable ( 1 3 ). Funama et al ( 1 ) and Nakayama et al ( 2 ) have shown that it is possible to reduce the tube voltage Radiology: Volume 257: Number 3 December 2010 n radiology.rsna.org 739

9 Figure 4 Figure 4: (a) Plot demonstrates a decrease in liver enhancement-to-noise ratio with the increase in patient longest linear dimension at the level of the liver by using 80-kV images. (b) Plot demonstrates an increase in image noise with increasing patient longest linear dimension at the level of the liver by using 80-kV images. settings from 120 to 90 kv in singlesource abdominal CT and still have acceptable image quality, but the patient size and noise level beyond which the image quality becomes unacceptable were not determined. Schindera et al ( 3 ) suggested the implementation of a lowtube-voltage CT technique (80 kvp) with a high tube current time product for the detection of hypervascular liver tumors during the arterial phase, but they specifically stated that additional research was needed to understand the effect of patient size on this practice. Our experience substantiates their important predictions relating to the possibility of dose reduction at 80 kv, as 76% (32 of 42) of the examinations performed with the mm detector configuration and 66% (49 of 74) of the examinations performed with the mm detector configuration had acceptable image quality at all anatomic levels and were performed with 50% or lower radiation dose (as the 80-kV image represented 50% of the radiation used). Further studies are needed to determine if low-tube-voltage imaging can be used to both lower radiation dose and improve or maintain diagnostic accuracy. Moreover, the cutoffs we have established are valid to our specific CTDI vol (in our practice the 80-kV tube CTDI vol is set at 50% or below the CTDI vol of the corresponding single-source 120-kV scan) and will be larger as the tube current or the section thickness increase. Our results are of benefit to users of single- and dual-source CT systems who are seeking to reduce dose and/or increase conspicuity of hypervascular and/or hypovascular lesions by using low-tube-voltage imaging ( ). According to the manufacturer and results of a phantom study ( 25 ), by using the on-line cross-scatter correction algorithm (which can only be used with the mm detector configuration), negligible differences in the CT number and noise are observed between single-source and dual-source images, independently of size and tube voltage settings. If only 80-kV imaging was used, a 50% dose reduction could have been achieved in many of our patients of appropriate size (as outlined by our size criteria). Furthermore, setting size criteria for dose reduction by using lower tube voltage imaging goes hand in hand with the development of multiple avenues of image noise reduction, such as adaptive statistical iterative reconstruction, projection space denoising, and other methods that will reduce noise and further improve image quality 740 radiology.rsna.org n Radiology: Volume 257: Number 3 December 2010

10 Table 3 Patient Size Cutoffs with Use of the Longest Linear Dimension (in Centimeters) Above Which a Sensitivity of 90% or More Can Be Achieved for Prediction of Unacceptable Image Quality Anatomic Level mm Detector mm Detector Liver 35.8 (0.94) 31.5 (0.90) Pancreas 34.9 (0.86) 33.0 (0.95) Kidneys 35.6 (0.66) 36.6 (0.94) Terminal ileum 35.0 (0.73) 35.2 (0.85) Note. Number in parentheses is the area under the receiver operating characteristic curve. Table 4 Cutoffs of Image Noise in Subcutaneous Fat Above Which a Sensitivity of 90% or More Can Be Achieved for Prediction of Unacceptable Image Quality Anatomic Level mm Detector mm Detector Liver 18 (0.75) 19 (0.84) Pancreas 20 (0.81) 22 (0.90) Kidneys 29 (0.93) 24 (0.92) Terminal ileum 27 (0.90) 24 (0.92) Note. Data are Hounsfi eld units. Number in parentheses is the area under the receiver operating characteristic curve. Table 5 Unacceptable Image Quality Cutoffs at the Level of the Liver by Using Clinical Measures of Patient Size with 90% or Greater Sensitivity Patient Size mm Detection mm Detector Weight (Kg).78.7 (0.89).67.4 (0.88) BMI.26.5 (0.87).24.1 (0.86) Body surface area (m 2 ).1.94 (0.86).1.76 (0.87) Note. Number in parentheses is the area under the receiver operating characteristic curve. of low-tube-voltage images ( ). In recent, retrospective multireader studies evaluating lesion conspicuity at hepatic and renal imaging, we have found that half-dose 80-kV CT data (such as the images evaluated herein) actually increased the conspicuity of hepatic and renal masses after noise reduction by using projection space denoising images, compared with full-dose mixedtube-voltage CT images ( 29,30 ). Our patient size cutoffs can therefore serve as the starting point for those seeking to perform single-source examinations at 80 kv as a means of radiation dose reduction. In our practice, we use the patient s longest linear dimension cutoffs of 36 and 41 cm for single-source 80-kV and 100-kV imaging, respectively. Another aspect that has to be taken into consideration when selecting patients for dual-source, dual-energy CT is the size of the smaller (26 cm) of the two detectors on the first-generation scanner (the one used in this experiment), which will prevent imaging of the entire field of view in larger patients. This field of view limitation will not be a problem for the aorta, pancreas, adrenal glands, or small bowel. When imaging the liver and kidneys, careful positioning within the field of view requires the acquisition of anteroposterior and lateral scout images. This difficulty is greatly reduced in the second-generation dual-source scanner, which has a smaller detector size of 33 cm. Our study had limitations. It was retrospective and had small number of cases with unacceptable image quality with use of the mm configuration at the levels of the kidneys and terminal ilium (due to a combination of patient acceptance and self-selection), which reduced the power of the statistical operations involving data from these two anatomic levels. Additionally, one might expect the quantitative analysis to match the qualitative analysis in the comparison of image acceptability produced by the two detector configurations. However, we find this result to be justifiable, since image quality is affected by factors other than noise, and the quantitative analysis cannot take these into account. Moreover, the noise levels that produced unacceptable image quality were virtually identical for both detector configurations, validating results of the qualitative analysis performed by the gastrointestinal radiologist. Because our results were from the first-generation dual-source scanner, our findings serve as a starting point for extension of appropriate patient size cutoffs for 80-kV imaging to other vendors and subsequent generations of dual-source CT scanners. We also evaluated image quality and artifacts, not diagnostic accuracy. Finally, we evaluated image quality during different phases of parenchymal enhancement at both detector configurations. We believed this approach mimics clinical practice in which all abdominal organs are evaluated during each phase of enhancement. Moreover, the image artifacts degrading low-tube-voltage images relate to patient size and the lower energy of the photon beam and should be independent of the iodine enhancement. If artifacts are minimal, then iodine-related image enhancement will be improved by lower tube voltage imaging. In conclusion, 80-kV abdominal CT can be performed with acceptable image quality when careful attention is paid to CT acquisition techniques, patient size, and organ of interest. The use of the mm detector configuration at Radiology: Volume 257: Number 3 December 2010 n radiology.rsna.org 741

11 abdominal dual-energy CT results in significantly better image quality than that obtained with the mm configuration. In the selection of patients who will benefit from 80-kV imaging, 35.8 cm at the level of the liver seems to be an appropriate patient size cutoff, but this value has to be modified according to the organ of interest and the desired CTDI vol. References 1. Funama Y, Awai K, Nakayama Y, et al. Radiation dose reduction without degradation of low-contrast detectability at abdominal multisection CT with a low-tube voltage technique: phantom study. Radiology 2005 ; 237 ( 3 ): Nakayama Y, Awai K, Funama Y, et al. Abdominal CT with low tube voltage: preliminary observations about radiation dose, contrast enhancement, image quality, and noise. Radiology 2005 ; 237 ( 3 ): Schindera ST, Nelson RC, Mukundan S Jr, et al. Hypervascular liver tumors: low tube voltage, high tube current multi-detector row CT for enhanced detection phantom study. Radiology 2008 ; 246 ( 1 ): Ende JF, Huda W, Ros PR, Litwiller AL. Image mottle in abdominal CT. Invest Radiol 1999 ; 34 ( 4 ): Huda W, Lieberman KA, Chang J, Roskopf ML. Patient size and x-ray technique factors in head computed tomography examinations. II. Image quality. Med Phys 2004 ; 31 ( 3 ): Nickoloff EL, Dutta AK, Lu ZF. Influence of phantom diameter, kvp and scan mode upon computed tomography dose index. Med Phys 2003 ; 30 ( 3 ): Szucs-Farkas Z, Verdun FR, von Allmen G, Mini RL, Vock P. Effect of X-ray tube parameters, iodine concentration, and patient size on image quality in pulmonary computed tomography angiography: a chest-phantomstudy. Invest Radiol 2008 ; 43 ( 6 ): Huda W, Lieberman KA, Chang J, Roskopf ML. Patient size and x-ray technique factors in head computed tomography examinations. I. Radiation doses. Med Phys 2004 ; 31 ( 3 ): Sigal-Cinqualbre AB, Hennequin R, Abada HT, Chen X, Paul JF. Low-kilovoltage multi- detector row chest CT in adults: feasibility and effect on image quality and iodine dose. Radiology 2004 ; 231 ( 1 ): Nakayama Y, Awai K, Funama Y, et al. Lower tube voltage reduces contrast material and radiation doses on 16-MDCT aortography. AJR Am J Roentgenol 2006 ; 187 ( 5 ): W490 W Flohr TG, McCollough CH, Bruder H, et al. First performance evaluation of a dualsource CT (DSCT) system. Eur Radiol 2006 ; 16 ( 2 ): Graser A, Johnson TR, Chandarana H, Macari M. Dual energy CT: preliminary observations and potential clinical applications in the abdomen. Eur Radiol 2009 ; 19 ( 1 ): Johnson TR, Krauss B, Sedlmair M, et al. Material differentiation by dual energy CT: initial experience. Eur Radiol 2007 ; 17 ( 6 ): Primak AN, Fletcher JG, Vrtiska TJ, et al. Noninvasive differentiation of uric acid versus non-uric acid kidney stones using dual-energy CT. Acad Radiol 2007 ; 14 ( 12 ): Warp RJ, Dobbins JT 3rd. Quantitative evaluation of noise reduction strategies in dual-energy imaging. Med Phys 2003 ; 30 ( 2 ): Fletcher JG, Takahashi N, Hartman R, et al. Dual-energy and dual-source CT: is there a role in the abdomen and pelvis? Radiol Clin North Am 2009 ; 47 ( 1 ): Siegel MJ, Schmidt B, Bradley D, Suess C, Hildebolt C. Radiation dose and image quality in pediatric CT: effect of technical factors and phantom size and shape. Radiology 2004 ; 233 ( 2 ): Mosteller RD. Simplified calculation of body-surface area. N Engl J Med 1987 ; 317 ( 17 ): Lam TK, Leung DT. More on simplified calculation of body-surface area. N Engl J Med 1988 ; 318 ( 17 ): Hsieh J. Adaptive streak artifact reduction in computed tomography resulting from excessive x-ray photon noise. Med Phys 1998 ; 25 ( 11 ): Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977 ; 33 ( 1 ): Guimaraes L, Fidler JL, Ehman EC, et al. Can we drop the 140 kv for hepatic mass lesion detection at contrast-enhanced, dualenergy hepatic CT? [abstr]. In: Radiological Society of North America scientific assembly and annual meeting program. Oak Brook, Ill: Radiological Society of North America, 2009 ; Macari M, Spieler B, Kim D, et al. Dualsource dual-energy MDCT of pancreatic adenocarcinoma: initial observations with data generated at 80 kvp and at simulated weighted-average 120 kvp. AJR Am J Roentgenol 2010 ; 194 ( 1 ): W27 W Robinson E, Babb JS, Chandarana H, Godoy MCB, Megibow AJ, Macari M. Dual-source dual-energy MDCT: comparison of 80 kvp and weighted average 120 kvp data for conspicuity of hypovascular liver metastases [abstr]. In: Radiological Society of North America scientific assembly and annual meeting program. Oak Brook, Ill: Radiological Society of North America, 2009 ; Eusemann C, Apel A, Schmidt B, et al. An on-line cross-scatter correction algorithm for dual-source CT (DSCT): effects on CT number accuracy and noise. In: Ehsan S, Hsieh J, eds. Proceedings of SPIE: medical imaging 2009 Physics of medical imaging. Vol Bellingham, Wash: SPIE The International Society for Optical Engineering, 2009 ; Hara AK, Paden RG, Silva AC, Kujak JL, Lawder HJ, Pavlicek W. Iterative reconstruction technique for reducing body radiation dose at CT: feasibility study. AJR Am J Roentgenol 2009 ; 193 ( 3 ): Thibault JB, Sauer KD, Bouman CA, Hsieh J. A three-dimensional statistical approach to improved image quality for multislice helical CT. Med Phys 2007 ; 34 ( 11 ): Manduca A, Yu L, Trzasko JD, et al. Projection space denoising with bilateral filtering and CT noise modeling for dose reduction in CT. Med Phys 2009 ; 36 ( 11 ): Ehman E, Guimaraes L, Fidler J, et al. Low kv imaging with projection space denoising to significantly reduce radiation dose and preserve lesion conspicuity and image quality at hepatic CT. Presented at the Abdominal Radiology Course of the Society of Gastrointestinal Radiologists, Orlando, Fla, February 21-26, Paulsen S, Takahashi N, Kawashima A, et al. Multi-reader evaluation of renal masses using low kv (80-kV) CT urography (CTU): potential for dose reduction and structural characterization using projection space denoising. Presented at the Abdominal Radiology Course of the Society of Gastrointestinal Radiologists, Orlando, Fla, February 21-26, radiology.rsna.org n Radiology: Volume 257: Number 3 December 2010