Jaydev K. Dave a) * Department of Radiology, Thomas Jefferson University, Philadelphia, PA 19107, USA

Transcription

1 Current state of practice regarding digital radiography exposure indicators and deviation indices: Report of AAPM Imaging Physics Committee Task Group 232 Jaydev K. Dave a) * Department of Radiology, Thomas Jefferson University, Philadelphia, PA 19107, USA A. Kyle Jones* Department of Imaging Physics, University of Texas MD Anderson Cancer Center, Houston, TX 77030, USA Ryan Fisher and Katie Hulme Department of Diagnostic Radiology, The Cleveland Clinic, Beachwood, OH 44122, USA Lynn Rill Department of Radiology, University of Florida, Jacksonville Beach, FL 32250, USA David Zamora Department of Radiology, University of Washington, Seattle, WA 98195, USA Andrew Woodward Division of Radiologic Science, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599, USA Samuel Brady Department of Diagnostic Imaging, St. Jude Children s Research Hospital, Memphis, TN 38105, USA Robert D. MacDougall Department of Radiology, Boston Children s Hospital, Boston, MA 02115, USA Lee Goldman Department of Radiology, Hartford Hospital, Hartford, CT 06102, USA Susan Lang and Donald Peck Department of Radiology, Henry Ford Health System, Detroit, MI 48202, USA Bruce Apgar Agfa Healthcare, Greenville, SC 29601, USA S. Jeff Shepard Department of Imaging Physics, University of Texas MD Anderson Cancer Center, Houston, TX 77030, USA Robert Uzenoff Fujifilm Medical Systems U.S.A., Inc., Stamford, CT 06902, USA Charles Willis Department of Imaging Physics, University of Texas MD Anderson Cancer Center, Houston, TX 77030, USA (Received 4 December 2017; revised 24 August 2018; accepted for publication 19 September 2018; published 17 October 2018) Beginning with the advent of digital radiography systems in 1981, manufacturers of these systems provided indicators of detector exposure. These indicators were manufacturer-specific, and users in facilities with equipment from multiple manufacturers found it a challenge to monitor and manage variations in indicated exposure in routine clinical use. In 2008, a common definition of exposure index (EI) was realized in International Electrotechnical Commission (IEC) International Standard Ed. 1, which also introduced and defined the deviation index (), a number quantifying the difference between the detector EI for a given radiograph and the target exposure index (EI T ). An exposure index that differed by a constant from that established by the IEC and the concept of the deviation index also appear in American Association of Physicists in Medicine (AAPM) Report No. 116 published in The AAPM Report No. 116 went beyond the IEC standard in supplying a table (Table II in the report of TG-116) titled Exposure Indicator Control Limits for Clinical Images, which listed suggested ranges and actions to be considered for each range. As the IEC EI was implemented and clinical data were gathered, concerns were voiced that the control limits published in the report of TG-116 were too strict and did not accurately reflect clinical practice. The charge of task group 232 (TG-232) and the objective of this final report was to investigate the current state of the practice for CR/DR Exposure and Deviation Indices based on AAPM TG 116 and IEC-62494, for the purpose of establishing achievable goals (reference levels) and action levels in e1146 Med. Phys. 45 (11), November /2018/45(11)/e1146/ American Association of Physicists in Medicine e1146

2 e1147 Dave et al.: Report of AAPM Task Group 232 e1147 digital radiography. Data corresponding to EI and were collected from a range of practice settings for a number of body parts and views (adults and pediatric radiographs) and analyzed in aggregate and separately. A subset of radiographs was also evaluated by radiologists based on criteria adapted from the European Guidelines on Quality Criteria for Diagnostic Radiographic Images from the European Commission. Analysis revealed that typical distribution was characterized by a standard deviation (SD) of with mean values substantially different from 0.0, and less than 50% of values fell within the significant action limits proposed by AAPM TG-116 ( ). Recommendations stemming from this analysis include targeting a mean value of 0.0 and action limits at 1 and 2 SD of the based on actual data of an individual site. EI T values, values, and associated action limits should be reviewed on an ongoing basis and optimization of values should be a process of continuous quality improvement with a goal of reducing practice variation American Association of Physicists in Medicine [ Key words: detector exposure, deviation index, digital radiography, exposure index, quality improvement, reference level 1. INTRODUCTION Beginning with the advent of digital radiography systems in 1981, manufacturers of these systems provided indicators of detector exposure. These indicators were manufacturerspecific, and users in facilities with equipment from multiple manufacturers found it a challenge to monitor and manage variations in indicated exposure in routine clinical use. In 2008, a common definition of exposure index (EI) was realized in International Electrotechnical Commission (IEC) International Standard Ed Medical electrical equipment Exposure index of digital X-ray imaging systems Part 1: Definitions and requirements for general radiography. 1 IEC also introduced and defined the deviation index (), a number quantifying the difference between the EI for a given radiograph and the target exposure index (EI T ). As defined, EI is proportional to the air kerma that the detector would have received under standard beam conditions for the same raw pixel value in the relevant image region. 1 The IEC Standard Ed defines the EI as: EI ¼ c 0 xgðvþ (1) with constant c 0 = 100 lgy 1, and g(v) representing air kerma in lgy at the image receptor corresponding to the values of interest (V) 1 and obtained from an equipment-specific inverse calibration function. An exposure index that differed by a constant from that established by the IEC and the concept of the deviation index also appear in AAPM Report No. 116, An Exposure Indicator for Digital Radiography, the report of American Association of Physicists in Medicine (AAPM) Task Group 116 published in ,3 IEC Standard Ed defines the as: EI ¼ 10 x log 10 ; (2) EI T with no specification for the precision of the reported and a footnote indicating that different EI T may be required for different examinations or applications. 1 AAPM Report No. 116 defines the as: ¼ 10 x log 10 K IND K TGTðb;vÞ ; (3) with one decimal of precision, K IND indicating the image receptor air kerma and K TGT (b,v) denoting target air kerma explicitly defined as a function of body part (b) and view (v). 2,3 Both EI and K IND represent the image receptor air kerma; however, EI as defined by the IEC is unitless and differs from K IND by a constant factor of 100 lgy 1. Other than the differences in specified precision, the deviation indices defined by IEC Ed. 1 and AAPM Report No. 116 are the same. AAPM Report No went beyond the IEC standard in supplying a table (Table II in the report of TG-116) titled Exposure Indicator Control Limits for Clinical Images, reproduced as Table I in this report. This table listed suggested ranges and actions to be considered for each range. As the IEC EI was implemented and clinical data were gathered, concerns were voiced that the control limits published in the report of TG-116 were too strict and did not accurately reflect clinical practice. Of particular concern was the recommendation to Consult radiologist for repeat for a less than 1.0. The charge of task group 232 (TG-232), as approved by the Science Council of the AAPM, was to investigate the current state of the practice for CR/DR Exposure and Deviation Indices based on AAPM TG 116 and IEC-62494, for the TABLE I. Table II from AAPM Report No. 116: exposure indicator control limits for clinical images a. Range action(s) >+3.0 Excessive patient radiation exposure Repeat only if relevant anatomy is clipped or burned out Require immediate management follow-up +1.0 to +3.0 Overexposure: Repeat only if relevant anatomy is clipped or burned out 0.5 to +0.5 Target range Less than 1.0 Underexposed: Consult radiologist for repeat Less than 3.0 Repeat a Note, these are not the recommendations of TG-232.

3 e1148 Dave et al.: Report of AAPM Task Group 232 e1148 purpose of establishing achievable goals (reference levels) and action levels in digital radiography. TABLE III. Mandatory and optional data criteria. Mandatory data criteria Optional data criteria 2. DATA COLLECTION This task group collected digital radiography EI data from a range of practice settings, including academic medical centers, large healthcare organizations, community clinics, and pediatric imaging departments. Data were collected for a limited number of body parts and views for adult and pediatric patients (Table II) to ensure adequate sampling and limit the scope of work for the task group. Individual task group members submitting data complied with the requirements of the Institutional Review Board or Quality Improvement Assessment Board at their site. Clinical EI data had to meet certain mandatory and optional data criteria to be included in the analysis (Table III). These criteria ensured data reliability and aided the task group in data analysis. The EI T foreachdatapointwasusedtocalculatethe using Eq. (2). For facilities that used manufacturer-specific indices, we converted the values based on known relationship between EI and image receptor air kerma. 2 Table IV contains a list of equipment manufacturers at the participating sites from which the data were acquired for this report. Methods for determining the EI T varied at different institutions, with some using one or more specific targets recommended by vendor applications specialists, a qualified medical physicist, or a radiologist; while others used EI T calculated based on a number of actual clinical radiographs. Some institutions used different EI T for different body parts and views, while others used a single EI T for all body parts and views. A compilation of these EI T values is provided for reference in Table V. Participants were asked to provide data for a period of 1 yr for as many scanned-pixel (i.e., computed radiography [CR]) digitizers and fixed-pixel (i.e., digital radiography [DR]) rooms as possible. 3. DATA ANALYSIS The collected data represented a diverse array of practice environments and digital radiography technology. Practice TABLE II. Body parts and views analyzed. AP abdomen (KUB a ) Upright abdomen Decubitus abdomen AP a chest PA a chest Lateral chest Decubitus chest AP pelvis Extremity b a KUB: kidney-ureters-bladder; AP: anteroposterior, PA: posteroanterior. b Extremity included any view of the arm from the humerus distal to the hand, and the leg from the knee distal to the foot. Studies in which multiple views were acquired on one image were excluded. 1. Anonymized and NO PHI a is included 2. Must be collected from systems with known relationship between EI and image receptor air kerma 3. EI calibration of vendor used, or welldefined alternative used 4. Known equipment problems disclosed and linked to suspect data 5. EI T known for each data point 6. Any processing of data disclosed 7. Body part and view listed for each data entry a PHI = protected healthcare information. b AEC = automatic exposure control. 1. Radiographic room where image was acquired 2. Beam quality for each radiographic room 3. Method of exposure control (AEC b /manual) known for each data point 4. Status of image (accepted/ rejected) and reason for rejection if known TABLE IV. Types of equipment and equipment manufacturers. Manufacturer Agfa (Agfa Healthcare USA, Greenville, SC) Carestream (Carestream Health, Rochester, NY) Fujifilm (Fujifilm, Valhalla, NY) GE (GE Healthcare, Waukesha, WI) Philips (Koninklijke Philips, Amsterdam, Netherlands) Siemens (Siemens Medical Solutions, Malvern, PA) Scanned-pixel ( CR ) equipment x x Fixed-pixel ( DR ) equipment TABLE V. EI T (IEC) values used at sites submitting data for analysis by this task group. Body part Adult/ pediatric Technology Number of participating sites x x x x x Range of EI T a Abdomen Adult Scanned pixel (CR) Fixed pixel (DR) Pediatric Scanned pixel (CR) Fixed pixel (DR) Chest Adult Scanned pixel (CR) Fixed pixel (DR) Pediatric Scanned pixel (CR) Fixed pixel (DR) Pelvis Adult Scanned pixel (CR) Fixed pixel (DR) Pediatric Scanned pixel (CR) Fixed pixel (DR) Extremity Adult Scanned pixel (CR) Fixed pixel (DR) Pediatric Scanned pixel (CR) Fixed pixel (DR) a Note the EI T values were provided by the participating sites and these values may or may not match values recommended by equipment manufacturers. At each participating site, there were no differences in the EI T values for different views of the same body part for the same imaging technology from the same manufacturer.

4 e1149 Dave et al.: Report of AAPM Task Group 232 e1149 environments included academic hospitals, large hospital systems, and community hospitals. Data from different sites were processed in slightly different ways as described in Table VI. Only image instances with were considered in the analysis, according to the definition of the by TG A total of 505,930 image instances met this criterion and were analyzed (Table VII). data for each body part and view were stratified for analysis as illustrated in Fig. 1. data were normally distributed (Fig. 2). Descriptive statistics including mean, 95% confidence interval (CI) about the mean, median, variance, standard deviations (SD), minimum, maximum, range, interquartile range, skewness, and kurtosis were calculated for each of the data groups (one data group was a body part and view for either adults or pediatric TABLE VI. EI and data processing by site. Site identifier Data processing steps A, B, E J No processing C 1. Images processed with incorrect menus removed a 2. Images rejected for technical reasons removed D 1. Images with L value b indicative of unexposed plate removed 2. Images processed with incorrect menus removed a a If different anatomical location or view was selected relative to an acquired radiograph. b The L value is specific to Fujifilm CR and represents the dynamic range of the digitized exposure data, L = 4.0 indicates an unexposed plate. TABLE VII. Breakdown of views analyzed by this task group. Body Part Adult/pediatric View Number of views % of total views Abdomen Adult KUB % Upright % Decubitus % Pediatric KUB/Babygram % Upright % Decubitus 208 <0.1% Chest Adult AP % PA % Lateral % Decubitus % Pediatric AP % PA % Lateral % Decubitus 15 <0.1% Pelvis Adult AP % Pediatric AP % Extremity Adult Included views a % Pediatric Included views a % a View of the arm from the humerus distal to the hand, and the leg from the knee distal to the foot. Studies in which multiple views were acquired on one image were excluded. patients at a single site) and for the aggregated data using SPSS (IBM, Armonk, NY). The full data analysis is presented in Appendix A in Data S1 to this report (Supporting Information). data for each body part and view were classified according to the original criteria of Table II from AAPM Report No. 116 to provide a snapshot of current clinical practice (Tables VIII and IX). It can be observed from Tables VIII and IX that the of many diagnostic radiographs fell outside the recommended range ( 0.5to0.5)proposedbyAAPMTG-116andwould, according to Table II from their report, require repeated imaging, radiologist consult, or management follow-up. Interpretation of this snapshot of clinical practice is complicated by the fact that for many data groups the mean of the distribution was not 0.0 (Fig. 3). However, the typical spread in the distribution (evident from the interquartile ranges in Fig. 3) was still characterized by a large fraction of exposures for which the fell outside the action limits proposed by AAPM TG-116.Thestateofclinicalpracticeobservedbythistask group indicated that different action limits were required. In order to set these new action limits, the task group first attempted to determine the values at which the diagnostic utility of a clinical image is reduced or compromised. The approach selected by this task group was to conduct an image review with practicing radiologists to determine values, both negative and positive, at which the diagnostic utility of clinical images is reduced. 4. IMAGE REVIEW The task group decided to evaluate 2 SD of the collected clinical data as a starting point for significant action limits. As can be observed from Table X, this starting point included rather large for some body parts and views. Note that the limits of investigation in Table X are asymmetric as they were selected as a starting point from the collected clinical data without scaling the mean to be equal to 0.0. The first matter to be addressed for the proposed action limits was to verify that all values of interest (VOI) would be contained within the dynamic range of digital image receptors at these values. A series of 45 FOR PROCESSING clinical images from a fixed-pixel indirect digital radiography system including the views PA chest (125 kvp, with grid), lateral chest (125 kvp, with grid), AP abdomen (80 90 kvp, with grid), and extremities (50 80 kvp, with or without grid) was reviewed to determine the input dynamic range of the data. The FOR PROCESSING images from this DR system had a linear characteristic function, therefore, the input dynamic range of the data was equal to the ratio of the maximum FOR PROCESSING pixel value to the minimum FOR PROCESSING pixel value. The corresponding approximate minimum and maximum air kerma delivered to the image receptor was calculated from EI T, the input dynamic range of the data, and the. According to the specifications of this digital radiography system, which were located in the operator s manual, the detector dynamic range (RQA5 beam

5 e1150 Dave et al.: Report of AAPM Task Group 232 e1150 (a) (b) FIG. 1. Stratification of data for analysis. (a) based on types of exposure control methodology and imaging practice; (b) based on types of exposure control methodology and image receptor technology. (AEC: automatic exposure control; CR: computed radiography [i.e., scanned pixel]; DR: digital radiography [i.e., fixed pixel]). quality 4 ), henceforth referred to as the latitude to avoid confusion with the input dynamic range of the body part and view imaged, spanned from 5.3 ngy to 68.3 lgy. Two input dynamic ranges were considered one containing all image data out to the edge of the skin line but excluding lead markers (input dynamic range), and one excluding the skin line ( practical input dynamic range ) (Fig. 4). The input dynamic ranges characterizing the body parts and views evaluated are listed in Table XI, along with the associated EI T selected for the views, based on the data collected for this report (Table V). As an example for the purposes of illustration, consider an extremity view with EI T = 1128 (K TGT = lgy under the calibration conditions for the EI). Assuming the EI represents the central tendency of the FOR PROCESSING pixel values, at a = 5.8 (Table X), corresponding to EI/ EI T = 10^(5.8/10) = 3.8, the maximum air kerma at the Not to be confused with exposure latitude, the range of mas values that can be used for the same patient and view while still obtaining a diagnostic quality image. FIG. 2. data for adult extremity views (from all sites; n = 181,674) illustrating the normal distribution of values.

6 e1151 Dave et al.: Report of AAPM Task Group 232 e1151 image receptor contained in the full input dynamic range (35) would likely exceed the latitude of the image receptor (Fig. 5). Therefore, the task group re-evaluated the proposed action limits and decided to use = 3.0 as the starting point for the image review in an effort to avoid data with true overexposure (i.e., saturation). The image review was conducted by having two boardcertified radiologists compare a prototypical normal image of a specific body part and view to a series of images with surrounding the proposed lower and upper limits using a set of objective criteria and a 3-point rating scale. PA chest, lateral chest, AP abdomen, AP pelvis, and extremity views were considered. The evaluation criteria used varied by body part and view and were adapted from the European Guidelines on Quality Criteria for Diagnostic Radiographic Images from the European Commission 5 and are included in Appendix B in Data S1 (Supporting Information) to this report. Images included in the review had to meet all of the screening criteria listed in Appendix B in Data S1 (Supporting Information). The prototypical normal image had to meet the same criteria and have a equal to All images included in the review were randomly selected from images meeting the specified criteria at participating sites and inspected against these criteria by the members of the task group who participated in the review, without restriction to the time period during which they were acquired. Image evaluation was conducted using a 3-point scale: 1 = Unacceptable, must be repeated 2 = Marginal 3 = Acceptable A single rating was assigned to each image based on the evaluation criteria. Radiologists were instructed to evaluate images as they normally would, including but not limited to adjusting window width and level and applying image processing available on the PACS system used. Radiologists were blind to the and EI of all images other than the prototypical normal images. TABLE VIII. Percentage of cases falling within the categories listed in Table II of the report of AAPM TG-116 for adult patients (the category with the maximum number of instances for each view is listed in bold). Abdomen Chest Body part Pelvis Extremity View KUB Upright Decubitus AP PA Lateral Decubitus AP Included views a n <-3 5.3% 11.1% 11.5% 5.6% 10.6% 4.0% 20.9% 4.9% 20.1% 3to % 17.1% 22.1% 18.0% 26.8% 22.7% 25.6% 11.9% 23.1% 1to % 7.0% 7.5% 8.2% 8.7% 9.9% 4.7% 7.0% 7.0% 0.5 to % 15.3% 15.1% 18.9% 17.1% 18.1% 10.1% 19.8% 13.9% 0.5 to 1 7.9% 8.4% 7.1% 9.8% 8.4% 7.9% 6.6% 11.3% 6.4% 1to3 31.5% 24.2% 19.3% 28.0% 23.0% 25.7% 21.7% 32.1% 17.7% >3 19.9% 17.0% 17.4% 11.5% 5.4% 11.8% 10.5% 13.0% 11.8% a View of the arm from the humerus distal to the hand, and the leg from the knee distal to the foot. Studies in which multiple views were acquired on one image were excluded. TABLE IX. Percentage of cases falling within the categories listed in Table II of the report of AAPM TG-116 for pediatric patients (the category with the maximum number of instances for each view is listed in bold). Abdomen Chest Body part Pelvis Extremity View KUB Upright Decubitus AP PA Lateral Decubitus AP Included views a n < % 32.6% 15.4% 10.7% 6.5% 4.0% 6.7% 10.4% 12.0% 3to % 33.6% 32.2% 25.6% 21.7% 11.7% 20.0% 20.5% 17.7% 1to % 2.3% 6.7% 9.6% 7.2% 7.4% 6.7% 10.2% 5.6% 0.5 to % 9.8% 14.4% 22.6% 21.7% 19.3% 20.0% 22.3% 13.4% 0.5 to 1 7.9% 2.6% 6.7% 6.7% 10.9% 11.1% 6.7% 8.2% 7.1% 1 to % 15.3% 20.2% 16.3% 26.1% 31.8% 33.3% 19.2% 26.3% >3 12.3% 3.9% 4.3% 8.5% 5.9% 14.7% 6.7% 9.1% 17.8% a View of the arm from the humerus distal to the hand, and the leg from the knee distal to the foot. Studies in which multiple views were acquired on one image were excluded.

7 e1152 Dave et al.: Report of AAPM Task Group 232 e1152 FIG. 3. Boxplots of values for all sites for adult a) AP abdomen (KUB) and b) PA chest views. Whiskers represent 1.5*IQR (interquartile range), circles represent values outside 1.5*IQR, and asterisks represent values outside 3*IQR. Reference lines are plotted for = 0.0 (solid), = 1.0 (short dash), and = 3.0 (long dash). (Quantitative details/descriptive statistics in Appendix A in Data S1 Supporting Information; for AP abdomen (KUB) refer to Section A.1 and for PA chest to Section A.3). TABLE X. Lower and upper limits corresponding to 2 standard deviations (SD) (for views of adult patients). Body part and view Exposure control Lower limit Upper limit PA chest AEC Lateral chest AEC AP abdomen AEC Extremity Manual AEC = automatic exposure control, Manual = manual exposure control. 4.A. Image review results A total of 90 images were included in the image review. Upon first conducting the image review using clinical images with = or = , all images were rated as 3 (Acceptable). Therefore, this final criterion for the image review was updated to require that images had 5.0 or 5.0. It is these results that are reported in this section, and these criteria that are listed in Appendix B in Data S1 (Supporting Information). Because only specific sites participated in the image review, only examinations of adult patients were included. Not all images included in the image review had a radiologist interpretation performed, some were rejected or repeated images that had not been archived to PACS. No saturation or burnout was observed in any of the images from any site used in the image review. For PA chest images with from 4.6 to 5.4 (EI T = 400), no image was rated less than 3. For AP chest images, which were nongrid bedside images, images with from 5.9 to +6.2 (EI T = 400) were all rated 2 or 3, none were rated unacceptable. All AP abdomen images with from 6.9 to +4.6 (EI T = 400) were rated 2 or 3. Extremity images with ranging from 7.7 to +8.4 (EI T = 400 or 1752) were all rated 2 or 3. AP pelvis images with from +2.7 to +3.6 (EI T = 876) were all rated 2 or 3. All digital radiographs evaluated in this limited image review were, in terms of exposure to the image receptor, acceptable to the participating radiologists for diagnosis. It is already known that the tolerance of radiologists for reading through noise is highly variable, and can be extremely high, particularly in the setting of pediatric imaging. Anecdotal feedback from the participating radiologists indicated that factors other than exposure appropriateness played a larger role in image acceptability. Positioning was the most important of these factors, and for many extremity images an image rating of 2 was solely the result of poor positioning, resulting in poor reproduction of fat pads and fat/muscle planes. These two observations may imply that exposure appropriateness is not the major limiting factor in image acceptability, and that radiologist intervention for questions of image quality is likely necessary only at extreme values of the. In fact, neither of these observations should come as a surprise as published reports on rejected image rates have shown that images rejected for positioning outnumber those rejected for under- or overexposure. 6 However, there are alternative interpretations of these results. The most important of these may be that EI T values that are currently used in clinical practice are too high when such extremely negative values result in images that are still clinically acceptable. Such high EI T values may have resulted from differences in the concept of speed class in different countries and failure to fully adapt speed class targets (i.e., EI T values) to new technology. It is also interesting to note that two different sites participating in the image review used extremely different EI T for extremity views for the same image receptor technology (400 vs 1752), therefore, images with moderate values on the order of 5.0 from either site would be considered appropriately exposed at the other site. It is also important to note that radiologists were not asked to comment on the appropriateness of the patient exposure, in fact, they were aware of neither the EI nor the of the images they were scoring.

8 e1153 Dave et al.: Report of AAPM Task Group 232 e1153 (a) (b) (c) FIG. 4. Illustration of full and practical input dynamic range for a PA chest radiograph. (a) FOR PROCESSING PA chest radiograph; (b) same image shaded corresponding to part (c) histogram of FOR PROCESSING pixel values with the practical input dynamic range shaded green, which excludes the subdiaphragm (shaded red) and the skin line (shaded white); the full input dynamic range includes the red, green, and white shaded areas, and excludes only the area beyond the end of the skin line (shaded purple) and any lead markers, if present. ] TABLE XI. Input dynamic range of clinical FOR PROCESSING image data and associated EI T. Body part and view EI T (K TGT a in lgy) Full input dynamic range of view 5. RELATIONSHIP BETWEEN AND EI T Practical input dynamic range of view PA chest 544 (5.44) Lateral chest 425 (4.25) AP abdomen 340 (3.40) Extremities ( ) a K TGT = target air kerma. Considering the results of the image review and the wide range of EI T observed by this Task Group, it is worthwhile to examine the relationship between EI T and the width of the distribution. The correlation between EI T and SD of the was calculated using Pearson s product moment correlation for adult AP chest, PA chest, lateral chest, and KUB abdominal views (Table XII), and these data are visualized in Fig. 6. Significance was determined using a two-tailed test. An individual site may have more than one data point per plot in Figs. 6(a) 6(d), as some sites used different EI T for the same body part and view depending on image receptor technology, location within the site, etc. The results are mixed, as the calculated correlations ranged from strong and significant to absent depending on body part and view. 6. WHAT TO DO WITH THESE RESULTS? For each body part and view, the task group tabulated the number of examinations at each participating site and calculated the SD of the for those examinations. These data were then used to identify the individual sites with the smallest and largest SD of the. Table XIII lists the results for adult examinations, and Table XIV for pediatric examinations. Complete SD data from all participating sites are provided in Appendix A in Data S1 (Supporting Information).

9 e1154 Dave et al.: Report of AAPM Task Group 232 e1154 FIG. 5. The histogram of an AP elbow extremity view using EIT = 1128 for = 0.0 (light gray with black outline) and for = 5.8 (in dark gray with no outline). When = 0.0, the practical dynamic range of clinical information corresponds to the latitude of the detector (upper end of image receptor latitude is marked by a dashed line). Although some pixel values exceed the latitude of the image receptor, these represent directly exposed area outside the patient anatomy and contain no clinically relevant information. However, at = 5.8 (histogram plotted in dark gray with no outline), the view may contain pixel values within the practical dynamic range that exceed the latitude of the image receptor. For the purposes of this illustration, the histogram for the image at = 5.8 was simulated, but the bin widths for both histograms are the same. TABLE XII. Pearson s product moment correlation (r) between EI T and standard deviation (SD) of the. Body part and view N a r P-value KUB abdomen AP chest PA chest <0.001* Lateral chest * a N = number of distinct EI T values analyzed. Multiple instances of the same EI T value may exist if multiple sites used the same EI T. *Significant at P = Considering the snapshot of current clinical practice provided by the review of data in this report, it is clear that the significant action limits originally proposed by TG-116 require rethinking. In the best case scenario, i.e., smallest SD, observed in these data, the action limits recommended by TG-116 suggest that a radiologist would be consulted for questions of image quality on more than 30% of digital radiographs, which is inefficient and unnecessary. The highest action levels recommended by TG-116, including immediate management review or repeating of an image, occur at of 3.0. However, as seen in the data collected by this task group, a of 3.0 or +3.0 is barely one SD for views of the abdomen and extremities. Therefore, a new approach to setting limits is required. A single best course of action is not clear based on the date collected and reviewed by this task group. What is clear is that any fixed set of limits are not suitable as a starting point, as evidenced both by the differences in the SD of distributions among different practice settings and body parts and views (Appendix A in Data S1 Supporting Information). Instead, limits should be tailored to individual radiography practices, and should consider factors such as the latitude (dynamic range) of image receptors used and EI T values for different body parts and views. For example, the upper limit might be lower for a high EI T compared to a low EI T. Therefore, this task group has decided to publish general recommendations along with factors to consider when setting significant action limits. Table XV lists recommended starting points for limits and associated actions based on SD of the, and not fixed. Such limits are not in conflict with the goal of maintaining a mean of 0.0, instead they allow for variation among different practice settings, as can be observed in the data collected and analyzed by this task group. The vision of this task group is that eventually these starting points should transition to a set of fixed limits specific to an individual radiography practice, with narrowing of the distribution of values becoming a continuous quality improvement effort. It should also be noted that a site will be striving to achieve two goals: acquiring clinical radiographs with within the target limits, and maintaining a mean of 0.0. Achieving the first goal will be difficult if the mean differs substantially from RECOMMENDATIONS FOR EI T AND EI T should be configured with due consideration given to the practice, e.g., adult vs pediatric; the characteristics of the interpreting radiologists, e.g., tolerance of image noise; body part and view; image receptor technology and performance characteristics; image processing algorithm used; beam qualities used for clinical imaging; and the method used for VOI identification. It follows that limits should be configured with attention paid to the same factors. Radiographs should never be repeated based on alone, instead the decision to repeat should be based on a determination of the appropriateness of the reported and a review of image quality. Given the latitude of digital image receptors and the practical input dynamic ranges of common radiographic views, and depending on the selection of EI T, even radiographs with extreme may still be clinically acceptable. Significant action for radiographs with < 0.0, including repeating of images or consulting with a radiologist, should be reserved for images that have a high likelihood of being nondiagnostic. This task group recommends a tiered review process in which the technologist supervisor is consulted first, followed by consultation with a radiologist if required. Over time, the circumstances that necessitate repeated imaging will be reinforced through periodic consultation between the technical staff and radiologists, which should reduce the frequency of radiologist consult. The initial stage of consultation with the technologist supervisor should include a reasonable verification that the reported value is valid. This review process need not be limited to questions of exposure appropriateness, but can also be applied to positioning and other technical errors. Overexposed radiographs should not be repeated unless there is saturation (i.e., clipping) of diagnostically

10 e1155 Dave et al.: Report of AAPM Task Group 232 e1155 FIG. 6. Bubble plots of standard deviation (SD) of the vs EIT. The size of the bubble represents the number of radiographic exposures represented by each data point. (a) KUB abdomen; (b) AP chest; (c) PA chest; (d) lateral chest. ] important anatomy that cannot be rectified by adjustment of image processing settings (Fig. 9). However, overexposed images with greater than 2 SD should be logged and tracked for periodic review and appropriate corrective action taken and documented. The practice snapshot captured in the data reviewed by this task group and presented in this report can be used as a starting point for setting both EI T and limits, and individual sites can use their own data to adjust these limits over time. It is also worthwhile to consider the possibility of establishing absolute limits for both EI and. Consider the detector latitude in the example presented in the previous paragraph, 23 ngy to 68.3 lgy. Based on the practical input dynamic ranges for specific body parts and views listed in Table XI, absolute limits on EI can be derived. Regardless of, any image with an EI falling outside the absolute EI limits of a site would trigger an immediate repeat, as this would indicate that some pixels within the practical input dynamic range of the image were exposed to air kerma levels below or above the latitude of the image receptor. As it cannot be determined from the data collected by this task group what might be a lower bound on the SD of the, it is not possible at this time to set an absolute limit for the. Further, this lower bound might change as technology and the practice of radiography evolves. Regardless, a well-controlled radiography practice may find that after several cycles of quality improvement their distribution becomes stable and no longer narrows. A reduction in the SD of the would no longer be a target for quality improvement, but the SD would still be monitored as part of the overall quality assurance program. While radiographs with outside 2 SD from such a practice would still, by definition, be exceptional for the practice, they may be acceptable in terms of overall and the threshold for significant action may be set higher than 2 SD of the for such a practice. For the purposes of radiation dose management, sites may also wish to review EI data. As the is calculated as the log of the ratio of the EI to EI T, a mean = 0.0 does not correspond to a mean EI equal to EI T. Instead, as the distribution of the is normal (Fig. 2), the mean EI will always be higher than EI T, even when the mean = A. Practice considerations The practice setting will influence configuration of both EI T and limits. Pediatric radiologists may be more willing to read borderline images to avoid repeats, which will affect the chosen limits. Selection of limits may also be different for practices with a radiologic technologist teaching component. Other important considerations include the influence of beam quality and VOI identification on the EI. Practices that frequently use beam qualities that differ substantially from the beam quality used to calibrate the EI should account for these differences, which may also depend on the equipment manufacturer, in the establishment of EI T tables and limits. It is also critically important that practices understand the method used to identify the VOI and calculate the EI from

11 e1156 Dave et al.: Report of AAPM Task Group 232 e1156 TABLE XIII. Standard deviation (SD) of the for adult radiography. Site with the smallest SD of the a Site with the largest SD of the a Body part View Number of exams SD of Number of exams SD of Abdomen KUB Upright Decubitus Chest AP PA Lateral Decubitus Pelvis AP Extremity Lower Extremity Upper Extremity : Insufficient sample size (data provided in Appendix A in Data S1 (Supporting Information) for reference). a Number of examinations from site was at least 10% of the total number of examinations from all sites. TABLE XIV. Standard deviation (SD) of the for pediatric radiography. Site with the smallest SD of the a Site with the largest SD of the a Body part View Number of exams SD of Number of exams SD of Abdomen KUB/Babygram Upright Decubitus Chest AP PA Lateral Decubitus Pelvis AP Extremity Lower extremity Upper extremity : Insufficient sample size (data provided in Appendix A in Data S1 (Supporting Information) for reference). a Number of examinations from site was at least 10% of the total number of examinations from all sites. FIG. 7. Fault tree for greater than +2 standard deviations (SD). TABLE XV. Recommended action limits and associated actions for the deviation index (). Possible action outside 1 standard deviation Log for possible review, tally number of occurrences for periodic review greater than +2 standard deviations See fault tree in Fig. 7 less than 2 standard deviations See fault tree in Fig. 8 a Note these limits may need to be adjusted based on EI T. the VOI for all radiography equipment used in their practice, and to realize that these methods may vary even among different platforms from the same manufacturer. 7.B. Body part and view The input dynamic range of the body part and view imaged affect the choice of limits relative to EI T. This consideration may also necessitate the choice of asymmetric

12 e1157 Dave et al.: Report of AAPM Task Group 232 e1157 limits. For example, if EI T is 200, the lower limit for significant action might be higher than 2 SD (and must be greater than any absolute lower EI limit), while the upper limit is set to 2 SD, as less latitude is available on the underexposure side. The data reviewed in this report indicate that the distribution varies with body part and view. Practices performing a limited range of radiography examinations may elect to use a single set of limits, while practices performing radiography examinations that encompass a wide range of body parts and views may elect to use different limits for different body parts and views. Implicit in the discussion of the impact of body part and view on EI T values and limits is consideration of the clinical task. Demanding clinical tasks, such as detecting subtle joint effusions, may require a more strict lower limit, while less demanding tasks, such as visualization of air/fluid levels in the abdomen, may allow for more flexibility on the lower limit. These are decisions that must be made through collaboration among the radiologist, qualified medical physicist, and radiologic technologist manager. 7.C. Image receptor technology and specifications The latitude of the image receptor affects the choice of limits relative to EI T in a fashion identical to the body part and view. The available bit depth of the images also affects the choice of limits, as images with higher bit depth require less latitude to reproduce the same VOI. An additional consideration for scanned-pixel (i.e., CR) imaging systems is plate-to-plate variation in sensitivity. EI and limits should be adjusted according to the expected minimum and maximum plate sensitivity in the stock of plates used for clinical imaging. FIG. 8. Fault tree for less than 2 standard deviations (SD). 7.D. Image processing algorithm Even if unsaturated FOR PROCESSING images are output by an image receptor, saturated, i.e., clipped, FOR PRE- SENTATION images can result during the remapping of FOR PROCESSSING gray values to FOR PRESENTATION gray values by image processing algorithms (Fig. 9). It is FIG. 9. (a) Initial appearance of lateral chest radiograph, demonstrating saturation ( burnout ) in the retrosternal region (arrow), (b) saturation verified by adjusting window width (WW = 1) and window level (WL = 0), (c) problem corrected by selecting a different grayscale LUT avoiding repeated imaging and re-exposure of the patient.

13 e1158 Dave et al.: Report of AAPM Task Group 232 e1158 FIG. 10. Mean by body part and view for adult radiography from each clinical site that contributed data for this analysis. ] important during the image review process (Fig. 5) to determine if a diagnostic image can be recovered by adjusting image processing settings. Proper selection of limits can ensure that FOR PROCESSING data with an appropriate dynamic range are input to the clinical image processing algorithm. 7.E. Quality assurance Regardless of the EI T and limits used, ongoing monitoring of is an essential part of the quality assurance process. 6,7 Investigation of consistently high may identify the need for equipment service or education of the technical staff. If images with consistently low are produced and accepted by the radiologist, downward adjustment of EI T may be appropriate. This should be viewed more as an ongoing process of continuous improvement, with less emphasis placed on setting EI T values and limits initially. Quality tools such as control charts may be helpful for determining if the initial configuration of EI T values and limits was appropriate, and manufacturer resources such as whitepapers and software features are available to aid in initial configuration of EI T values and limits. Ongoing monitoring should include comparison of dose metrics to published normative data including reference levels and achievable levels, 6,7 and trending of and/or EI data over time to identify dose creep. 8,9 EI T should be reviewed on a regular basis, and updated based on findings from ongoing review of and EI data. 7.F. Tools to manage EI data It is essential that EI T values be set appropriately if values are to be used for quality control and quality improvement. Manufacturers of digital radiography equipment and manufacturers of radiation dose index monitoring systems (RMs) can help the clinical team in management of EI data during initial configuration and after hardware or software updates, including EI T settings and data, by providing simple tools. Such tools might include: Utilities for configuring global EI T values for broad categories, instead of requiring that an EI T value be configured for each body part and view for all patient classes. The ability to set limits at any level of granularity, from a single universal set of limits to limits by individual body parts and views. Both of the above may be accomplished by allowing upload of EI T values and limits in a specified file format. Utilities for easily filtering and downloading EI T values, EI data, and data, preferably over the network. An optional overlay of the identified VOI on the FOR PROCESSING image data. 7.G. Best practices This report has focused on the distributions that are actually encountered in typical radiology departments. However, this may not be indicative of what distributions are achievable in a well-managed radiology department with a focus on quality. It is interesting to note from the practice data analyzed by this task group that in many cases the variance in was virtually the same for both automatic and manual exposure control of the same body part and view (Appendix A in Data S1 Supporting Information). This can be interpreted in two ways: either the distribution of from images acquired using AEC represents best practice, and most practices that contributed data to this task group are doing about as well as could be expected for manual exposures; or, room for improvement exists even in AEC exposures. Defining expected distributions for data that represent best practice requires careful study of the effect of VOI identification, patient positioning, and other factors on the reported and its distribution.

14 e1159 Dave et al.: Report of AAPM Task Group 232 e1159 Ultimately, the choice of limits and the strategy used to manage them rest with the individual practice, and require the input of the radiologist, lead technologist, and qualified medical physicist. Ideally, practices would use data specific to their operation to set limits, and would work toward narrowing the limits as much as practical, eventually transitioning to a fixed set of limits. This approach, however, may not be practical for all radiography operations. Some practices may elect to use a fixed set of limits, or even a single set of universal limits, based on the data presented in this report. Practices using the data from this report to set initial or fixed limits should keep in mind that the data in Tables XIII and XIV reflect only the particular sites with the smallest and largest SD of the. The data in these two tables do not necessarily reflect the heterogeneity of the distribution across all participating sites and body parts and views. The complete data analysis from all participating sites is available in Appendix A in Data S1 (Supporting Information). Whatever the approach chosen, ongoing review of limits is advised. Future advances in VOI recognition and EI calculation may result in a convergence of disparate limits to a single set for all body parts and view. The opposite could occur as well. Finally, patient exposure is an important consideration. This report has focused mainly on limits as they relate to image quality and the ability of the radiologist to interpret an image. While a clinical image with a > 0.0 should never be repeated if the necessary clinical information is appropriately rendered, practices may wish to set stricter limits for corrective action for radiographs with unnecessarily high. The radiologist should be involved in these QI efforts; however, it is not necessary to consult the radiologist for each individual radiograph, instead, summary results can be reviewed during periodic meetings of the QI committee. 8. LIMITATIONS It is important to acknowledge the limitations of the work done by this task group and the limitations of the EI and in general. The clinical data analyzed in this report were collected from a number of different sites with different practices. Some sites had taken steps to create tables of EI T considering their unique practices, while some sites used manufacturer default values. Calculated EI, and therefore, values depend, sometimes in ways that are not well understood, on acquisition parameters, VOI identification, image processing settings, patient positioning, and other factors. EI values reported by the various systems from which data were collected for this analysis were not verified individually, instead, the Task Group relied on sites submitting data to verify that EI calibration conditions and reported EI values were accurate. Finally, mean values are meaningful only when appropriate EI T values have been selected and configured. It was evident from the data analyzed by this task group that the appropriateness of EI T values used by participating sites varied, even by body part and view (Fig. 10). It is also important to reiterate that the results of the image review study should not be used beyond how they have been considered in this report, as only a few radiologists reviewed a small sample of images. 9. FUTURE WORK The most immediate needs related to the clinical use of the are the establishment of a reasonable set of EI T values for common body parts and views, including methods for establishing EI T for clinical examinations for which the beam quality differs substantially from the EI calibration conditions, and a commitment from clinical practices to analyze data and act on these data to maintain appropriate EI T. Further study of the impact of acquisition and image processing settings, patient positioning, beam quality, and VOI identification on the EI is also needed, along with recommendations for EI T based on imaging task, patient characteristics, and body part and view. 10. TAKE-HOME POINTS The typical distribution of the deviation index () was characterized by a SD of , and is affected by a number of factors specific to individual practice settings and related to how the exposure indicator (EI) is defined and calibrated and values of interest (VOI) are identified. Many mean values for different sites and body parts and views differed substantially from 0.0, indicating that either target exposure indicator (EI T ) values were not set appropriately or that the technical staff were not properly exposing images. Even for a mean = 0.0, for a typical radiography practice, less than 50% of values fell within the significant action limits proposed by AAPM TG-116 ( ). Considering the input dynamic range of common body parts and views and the latitude of modern digital image receptors, significant action, including consulting with radiologists or repeating images, is necessary only at extreme values. This task group recommends that a mean of 0.0 be targeted for all body parts and views. This requires that EI T values be set appropriately. Manufacturers can aid the clinical team in managing EI data by providing simple tools on digital radiography systems. As a starting point, this task group recommends that action limits for the be set at 1 and 2 SD of the based on actual data of an individual site as detailed in this report. These limits should be reviewed periodically and eventually transitioned to a set of practice-specific fixed limits. Data from this report (Tables XIII and XIV) can be used as a starting point until sufficient data have been accumulated to create site-specific limits.