William P. Shuman, MD Keith T. Chan, MD Janet M. Busey, MS Lee M. Mitsumori, MD Eunice Choi, BS Kent M. Koprowicz, MS Kalpana M. Kanal, PhD.

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 William P. Shuman, MD Keith T. Chan, MD Janet M. Busey, MS Lee M. Mitsumori, MD Eunice Choi, BS Kent M. Koprowicz, MS Kalpana M. Kanal, PhD Standard and Reduced Radiation Dose Liver CT Images: Adaptive Statistical Iterative Reconstruction versus Model-based Iterative Reconstruction Comparison of Findings and Image Quality 1 Purpose: To investigate whether reduced radiation dose liver computed tomography (CT) images reconstructed with modelbased iterative reconstruction (MBIR) might compromise depiction of clinically relevant findings or might have decreased image quality when compared with clinical standard radiation dose CT images reconstructed with adaptive statistical iterative reconstruction (ASIR). Original Research n Gastrointestinal Imaging 1 From the Department of Radiology, University of Washington School of Medicine, Box , 1959 NE Pacific St, Seattle, WA Received March 20, 2014; revision requested May 5; final revision received June 4; accepted June 13; final version accepted June 18. Supported by an unrestricted grant from GE Healthcare. Address correspondence to W.P.S. ( wshuman@u. washington.edu). q RSNA, 2014 Materials and Methods: Results: Conclusion: With institutional review board approval, informed consent, and HIPAA compliance, 50 patients (39 men, 11 women) were prospectively included who underwent liver CT. After a portal venous pass with ASIR images, a 60% reduced radiation dose pass was added with MBIR images. One reviewer scored ASIR image quality and marked findings. Two additional independent reviewers noted whether marked findings were present on MBIR images and assigned scores for relative conspicuity, spatial resolution, image noise, and image quality. Liver and aorta Hounsfield units and image noise were measured. Volume CT dose index and size-specific dose estimate (SSDE) were recorded. Qualitative reviewer scores were summarized. Formal statistical inference for signal-to-noise ratio (SNR), contrastto-noise ratio (CNR), volume CT dose index, and SSDE was made (paired t tests), with Bonferroni adjustment. Two independent reviewers identified all 136 ASIR image findings (n = 272) on MBIR images, scoring them as equal or better for conspicuity, spatial resolution, and image noise in 94.1% (256 of 272), 96.7% (263 of 272), and 99.3% (270 of 272), respectively. In 50 image sets, two reviewers (n = 100) scored overall image quality as sufficient or good with MBIR in 99% (99 of 100). Liver SNR was significantly greater for MBIR ( [standard deviation] vs , P,.001); there was no difference for CNR ( vs , P =.45). For ASIR and MBIR, respectively, volume CT dose index was 15.2 mgy versus 6.2 mgy 6 3.6; SSDE was 16.4 mgy versus 6.7 mgy (P,.001). Liver CT images reconstructed with MBIR may allow up to 59% radiation dose reduction compared with the dose with ASIR, without compromising depiction of findings or image quality. q RSNA, 2014 Radiology: Volume 273: Number 3 December 2014 n radiology.rsna.org 793

2 For a number of decades, filtered back projection (FBP) has been the most widely used image reconstruction technique for computed tomography (CT). While FBP is common because it is fast, it results in relatively noisy images. Recently, several iterative reconstruction techniques have allowed for substantially lower image noise compared with that with FBP when images are reconstructed from the same CT data. Adaptive statistical iterative reconstruction (ASIR) is a partially iterative technique that lowers image noise below that of FBP by iteratively repeating back projection multiple times (1). The noise reduction from ASIR may be used to obtain better image quality than reconstruction with FBP or may be used for lower radiation dose scanning and comparable image quality. Model-based iterative reconstruction (MBIR) is a fully iterative technique that further reduces image noise through iterative use of both back and forward projections (1). Noise reduction from MBIR reconstruction may either result in better image quality than reconstruction with ASIR or may also be used for lower radiation dose scanning and comparable image quality (2 6). The appearance of CT images reconstructed with MBIR may be different in ways radiologists have reported as waxy or blotchy, which reflects the Advances in Knowledge nn Portal venous phase liver CT images with 59% reduced radiation dose reconstructed with model-based iterative reconstruction (MBIR) depicted all 136 clinically relevant findings identified on the standard radiation dose portal venous CT images reconstructed with adaptive statistical iterative reconstruction (ASIR). nn Subjective scores and objective measurements of image quality were similar or better with 59% reduced radiation dose MBIR images compared with standard radiation dose ASIR images (sizespecific dose estimate, 6.7 mgy vs 16.4 mgy). complexity of describing noise in CT images (7 9). The purpose of this study was to investigate whether reduced radiation dose CT images reconstructed with MBIR might compromise depiction of clinically relevant findings in the liver or might decrease image quality when compared with our clinical standard radiation dose CT images reconstructed with ASIR. We designed an imaging protocol that had two closely timed CT passes through the liver during the portal venous phase. The first pass was for our clinical standard radiation dose images reconstructed with ASIR. The second pass was for 60% reduced radiation dose images reconstructed with MBIR. Materials and Methods This prospective single-institution study was Health Insurance Portability and Accountability Act compliant and was approved by our institutional review board. Written informed consent was obtained from all patients. Patient Selection The inclusion criterion was that a patient had undergone prior CT that showed findings of advanced cirrhosis and was scheduled for clinically indicated follow-up multiphase liver CT. Clinical exclusion criteria were severe allergy to iodinated contrast material, compromised renal function (glomerular filtration rate of less than 40 ml/ min/1.73 m 2 ), pregnancy, body mass index more than 35 kg/m 2, or age younger than 35 years. Of potential research subjects, 15 were excluded for increased body mass index and six declined after reviewing the informed consent. Between April and August of 2012, we enrolled a sequential series of 50 patients (mean age, 57 years 6 7 Implication for Patient Care nn MBIR may allow substantial radiation dose reduction in liver CT without compromising depiction of clinically relevant findings and with acceptable or improved image quality. [standard deviation]; body mass index, 29 kg/m 2 6 4; 11 women [mean age, 57 years 6 10; range, years], 39 men [mean age, 57 years 6 6; range, years]). CT Scanning Technique All patients underwent multiphase 64 detector row liver CT (Discovery CT750 HD; GE Healthcare, Waukesha, Wis). Circulation time from the antecubital fossa to the abdominal aorta at the level of the celiac artery was estimated by using a timing bolus of 15 ml of intravenous contrast material (Omnipaque 350; GE Healthcare, Chalfont St Giles, England) followed by 15 ml of saline administered through a dual-head power injector (Stellant D; Medrad, Warrendale, Pa) at 5 ml/sec. Scanning was performed during injection of 150 ml of contrast material at a rate of 5 ml/sec. The multiphase CT consisted of four passes: one pass in the late arterial phase by using a clinical standard radiation dose, two passes in the portal venous phase (one by using a clinical standard radiation dose and one immediately after using a reduced radiation dose), and one final pass in the delayed Published online before print /radiol Content codes: Radiology 2014; 273: Abbreviations: ASIR = adaptive statistical iterative reconstruction CNR = contrast-to-noise ratio FBP = filtered back projection MBIR = model-based iterative reconstruction SNR = signal-to-noise ratio SSDE = size-specific dose estimate Author contributions: Guarantors of integrity of entire study, W.P.S., K.T.C., K. M. Kanal; 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 final version of submitted manuscript, all authors; agrees to ensure any questions related to the work are appropriately resolved, all authors; literature research, W.P.S., J.M.B., E.C.; clinical studies, W.P.S., K.T.C., J.M.B., L.M.M., E.C.; experimental studies, K.T.C., E.C.; statistical analysis, W.P.S., K.T.C., J.M.B., K. M. Koprowicz; and manuscript editing, W.P.S., K.T.C., J.M.B., L.M.M., K. M. Koprowicz, K. M. Kanal Conflicts of interest are listed at the end of this article. 794 radiology.rsna.org n Radiology: Volume 273: Number 3 December 2014

3 Figure 1 Figure 1: Chart shows timing of multiphase four-pass liver CT: one pass in the late arterial phase at standard radiation dose, two passes in the portal venous phase (one at standard radiation dose and one immediately after at reduced radiation dose), and one final pass in the delayed phase at standard radiation dose. The images from the standard dose portal venous pass were reconstructed with 40% ASIR, and the images from the reduced dose portal venous pass were reconstructed with MBIR. NI = noise index (expressed in Hounsfield units). phase by using a clinical standard radiation dose (Fig 1). The start of scanning for the late arterial phase began 20 seconds after aortic peak attenuation on the timing bolus. Subsequent scanning was performed at 65 and 72 seconds for portal venous phase passes and at 300 seconds for the delayed phase pass. Scanning for each pass was performed during relaxed inspiration in the craniocaudal direction from just above the diaphragm to the top of the iliac crests. For all passes, the following were utilized: a tube voltage of 120 kvp and a variable tube current (milliamperes) with automated tube current modulation based on a noise index of 41 HU for clinical standard radiation dose passes and a noise index of 66 HU for the reduced radiation dose pass. Examination parameters for the two portal venous phase passes are presented in Table 1. Late arterial phase and delayed phase passes were not evaluated in this investigation. Postprocessing and Image Reconstruction The images obtained with the clinical standard radiation dose first pass from the portal venous phase were reconstructed by using 40% ASIR (ASiR, GE Healthcare, Milwaukee, Wis) blended with 60% FBP, with reconstruction typically requiring 1 2 minutes. The images obtained with the reduced radiation dose second pass from the portal venous phase were reconstructed using MBIR (VEO; GE Healthcare, Milwaukee, Wis), with reconstruction typically requiring minutes. The mm contiguous sections from both portal venous passes were pushed to a postprocessing workstation where the images were reformatted into 2.5-mm axial contiguous sections and were deidentified (Advantage Workstation 4.6; GE Healthcare, Milwaukee, Wis). Fifty image sets were created, with each image set containing two series: Series one was the image set obtained with clinical standard radiation dose portal venous pass images reconstructed with ASIR, and series two was the image set obtained with reduced radiation dose portal venous pass images reconstructed with MBIR. All image sets were then pushed to a research folder on a picture archiving and communication system (Centricity PACS; GE Healthcare, Chicago, Ill). Image Evaluation Three reviewers working independently at separate times systematically reviewed image sets randomly presented over a 10-week period. Two of the reviewers were body imaging fellowshiptrained attending radiologists with 16 and 30 years of experience in abdominal CT (reviewers A [L.M.M.] and C [W.P.S.]), while one reviewer was a resident with 4 years of experience (reviewer B [K.T.C.]). Reviewers initially received standardized instructions and were trained on image sets from five patients not included in this study. Images were presented to reviewers with a window width of 400 HU and window level of 40 HU, but reviewers could vary the window/level at will. No time limits were placed on the image review process. On the picture archiving and communication system workstation, reviewer A looked only at the clinical standard radiation dose ASIR series and placed permanent arrows on all liver findings which normally would be reported in a routine clinical radiology report. Each liver finding marked with an arrow was assigned to one of 16 predefined categories by this reviewer. This reviewer then scored overall image quality of the ASIR series on a four-point scale: score 1, not evaluable because of very high image noise and marked distortion of spatial or contrast resolution unwilling to read in any clinical situation; score 2, poor because of high image noise and moderate distortion of spatial or contrast resolution willing to read only in certain clinical situations; score 3, fair with some compromise because of increased image noise and some distortion of spatial or contrast resolution willing to read in most clinical situations; and score 4, good quality comparable to routine clinical scanning with some standard image noise and very little distortion of spatial or contrast resolution willing to read in all clinical situations. This same reviewer then assigned a score for contour definition and structure delineation of in-plane blood vessels plus the liver margin on a four-point scale: score 1, severe blurring, edge definition very poor, margins difficult to discern; score 2, moderate blurring, edge definition poor, margins can be discerned; score 3, minimal blurring, edge definition good, margins easily discerned; score 4, no blurring, edges well defined, margins crisp. Working independently at different times, reviewers B and C looked at images from both the clinical standard radiation dose ASIR series (with findings on images previously marked with arrows by reviewer A) and the unmarked reduced dose MBIR series side by side for each of the 50 cases. These reviewers were asked if it were possible to identify the ASIR finding marked with an arrow on the images from the reduced dose MBIR series. They were next asked to agree or disagree with the categorization of the Radiology: Volume 273: Number 3 December 2014 n radiology.rsna.org 795

4 Table 1 Portal Venous Phase CT Examination and Postprocessing Parameters Parameters Standard Dose Pass Reduced Dose Pass Noise index (HU)* Detector collimation (mm) Scan field of view (cm) Reconstruction section thickness (mm) Reconstruction section interval (mm) Pitch 1.375: :1 Gantry rotation time (sec) Tube voltage (kvp) Tube current control ATCM ATCM Tube current range (ma) Reconstruction filter Standard Standard Reconstruction algorithm ASIR MBIR Note. ATCM = automatic tube current modulation. * With mm section thickness. finding by reviewer A. On images in the reduced dose MBIR series, these two reviewers then subjectively scored relative conspicuity of each finding, relative spatial resolution of each finding, and relative image noise of each finding as: (a) a lot less than on images with standard dose ASIR, (b) a little less than images with standard dose ASIR, (c) the same as images with standard dose ASIR, (d) a little more than images with standard dose ASIR, or (e) a lot more than images with standard dose ASIR. Using the same scales as reviewer A, the other reviewers scored overall image quality and contour-definition and structure-delineation on images of both the standard dose ASIR series and the reduced dose MBIR series. The senior reviewer (reviewer C) obtained and averaged three Hounsfield unit attenuation measurements by using a region of interest of at least 4 cm 2 in the right lobe of the liver in each series. Each liver region of interest was carefully placed in a separate relatively homogeneous area of liver parenchyma away from discernible vessels or focal changes in attenuation. Hounsfield units of attenuation of the aorta were measured and averaged in a similar fashion at the level of the superior mesenteric artery. This reviewer also obtained and averaged three measurements of image noise (defined as the standard deviation of the Hounsfield units) in subcutaneous fat of the anterior abdominal wall away from artifact or vessels. Signal-to-noise ratio (SNR) in the liver and contrast-to-noise ratio (CNR) between the liver and the aorta were calculated by using the following formulas: SNR = ROI L /N F and CNR = (ROI A 2 ROI L )/N F, where N F is noise in fat, ROI L is region of interest in liver, and ROI A is region of interest in aorta. The displayed volume CT dose index was recorded for both the standard radiation dose and reduced radiation dose portal venous phase passes. Sizespecific dose estimate (SSDE) was calculated by using the sum of the anterior-posterior and lateral dimensions at the level of the mid liver (10). All scores and measurements were recorded in pen on a case report form and entered into an electronic database (Access 2010; Microsoft, Redmond, Wash). Statistical Analysis Data were analyzed by using software (Excel 2010, Microsoft; SAS software, version 9.3, SAS Institute, Cary, NC). Categorical data (liver findings and image quality scores) were summarized by using counts and proportions. Continuous measures (image noise, SNR, CNR, volume CT dose index, and SSDE) were summarized with means Table 2 Categories of 136 Liver Findings on Images in ASIR Series Findings Reviewer A Hyperattenuating focus Well defined 2 (1.5) Ill defined 5 (3.7) Hypoattenuating focus Well defined 47 (34.6) Ill defined 31 (22.8) Focal calcification 7 (5.1) Perfusion abnormality 3 (2.2) Fatty infiltration 2 (1.5) Fibrosis, focal 5 (3.7) Nodular contour 4 (2.9) Thrombosis of portal vein 8 (5.9) Iodized oil deposition* 8 (5.9) Heterogeneous parenchyma 5 (3.7) Bile duct dilatation 4 (2.9) Hepatic vein thrombosis 1 (0.7) TIPS with internal hyperplasia 2 (1.5) Recanalized umbilical vein 2 (1.5) Note. Numbers are frequencies of scores. Percentages of totals are in parentheses, and percentages were rounded. TIPS = transjugular intrahepatic portosystemic shunt. * Lipiodol; Guerbet, Bloomington, Ind. 6 standard deviations. Paired t tests with Bonferroni corrections were used to investigate differences for SNR, CNR, volume CT dose index, and SSDE between the images obtained with standard radiation dose and reduced radiation dose portal venous phase passes. A P value of less than.05 was considered to indicate a significant difference. Results Reviewer A identified, marked with arrows, and categorized 136 liver findings on images from the 50 clinical standard radiation dose ASIR series (Table 2). This reviewer assigned a score for overall image quality as fair or good in 98% of the 50 images in the ASIR series, and none of the images were nonevaluable. This reviewer also assigned a score for contour definition and structure delineation as minimal blurring or no blurring in 80% of the images in the ASIR series and as 796 radiology.rsna.org n Radiology: Volume 273: Number 3 December 2014

5 Table 3 Figure 2 Reader A: Subjective Image Quality Scores for Standard Dose ASIR Score Value Overall image quality 1, nonevaluable 0 2, poor 1 (2) 3, fair and sufficient 25 (50) 4, good 24 (48) Contour and delineation 1, severe blurring 0 2, moderate blurring 10 (20) 3, minimal blurring 27 (54) 4, no blurring 13 (26) Note. Numbers are frequencies of scores, with n = 50. Percent ages of totals are in parentheses, and percentages were rounded. having moderate blurring in 20% of the images (Table 3). Reviewers B and C separately identified all of the findings marked with an arrow from the clinical standard radiation dose ASIR images on the reduced radiation dose MBIR images viewed side by side and agreed with each of the finding categorizations from reviewer A. They each subjectively assigned a score for the conspicuity of 94% of the findings marked with an arrow as the same or greater with MBIR than with ASIR (mean, 94.1% [256 of 272]) (with two reviewers, n = 272 findings and n = 100 image sets) (Figs 2, 3). They subjectively assigned a score for spatial resolution as the same or better with MBIR than with ASIR in 95% and 98% of the findings marked with an arrow. They each subjectively assigned a score for noise as the same or lower with MBIR than with ASIR in 99% of the images with findings marked with an arrow (Table 4). For subjective overall image quality scores, reviewers B and C assigned a score in 98% and 100% as fair or good for images in both the ASIR and MBIR series, and no image set was scored nonevaluable. Reviewer B assigned a score for overall image quality as the same between ASIR and MBIR in 45 image sets; of the remaining five image sets, MBIR was assigned a score that Figure 2: CT images (120 kvp, automatic tube current modulation) in a 63-year-old man with history of hepatocellular carcinoma. Left: Portal venous phase CT image obtained with standard clinical radiation dose and reconstructed with 40% ASIR (first pass) shows a well-defined hypoattenuating focus (arrow), marked by reader A. Right: Portal venous phase CT image obtained with 60% reduced radiation dose and reconstructed with MBIR (second pass) shows the same finding. Figure 3 Figure 3: CT images (120 kvp, automatic tube current modulation) in a 67-year-old man with history of treated hepatocellular carcinoma. Left: Portal venous phase CT image obtained with standard clinical radiation dose and reconstructed with 40% ASIR (first pass) shows thrombosis in the portal vein (arrow), marked by reader A. Right: Portal venous phase CT image obtained with 60% reduced radiation dose and reconstructed with MBIR (second pass) shows the same finding. was higher in three and ASIR was assigned a score that was higher in two. Reviewer C assigned a score for overall image quality as the same in 36 image sets; of the remaining 14 image sets, MBIR was assigned a score that was higher in all. For contour-definition and structure-delineation, readers B and C assigned a score for both ASIR and MBIR as minimal or as having no blurring in all image sets (Table 5). The measured noise in subcutaneous fat determined by reviewer C was significantly greater for the standard radiation dose ASIR pass images than for the low radiation dose MBIR pass images, while SNR was significantly lower. The measured CNR was the same for images with both passes (Table 6). The volume CT dose index for standard dose ASIR images was 15.2 mgy and for reduced dose MBIR images was 6.2 mgy (P,.001). The calculated mean SSDE for standard dose ASIR images was 16.4 mgy and for reduced dose MBIR images was 6.7 mgy 6 3.1, a 59% radiation dose reduction (range, 40% 71%). Radiology: Volume 273: Number 3 December 2014 n radiology.rsna.org 797

6 Table 4 Readers B and C: Relative Image Quality Scores of 136 Liver Findings Score Reader B Reader C Relative conspicuity of MBIR 1, a lot less conspicuous than ASIR 1 (0.7) 0 2, a little less conspicuous than ASIR 7 (5.1) 8 (5.9) 3, same as ASIR 111 (81.6) 100 (73.5) 4, a little more conspicuous than ASIR 17 (12.5) 28 (20.6) 5, a lot more conspicuous than ASIR 0 0 Relative spatial resolution of MBIR 1, a lot less well defined than ASIR 0 0 2, a little less well defined than ASIR 7 (5.1) 2 (1.5) 3, same as ASIR 117 (86.0) 126 (92.6) 4, a little more well defined than ASIR 12 (8.8) 8 (5.9) 5, a lot more well defined than ASIR 0 0 Relative image noise of MBIR 1, a lot noisier than ASIR 0 0 2, a little noisier than ASIR 1 (0.7) 1 (0.7) 3, same as ASIR 96 (70.6) 81 (59.6) 4, a little less noisy than ASIR 39 (28.7) 54 (39.7) 5, a lot less noisy than ASIR 0 0 Note. Numbers are frequencies of scores. Percentages of totals are in parentheses, and percentages were rounded. Standard radiation dose images reconstructed with ASIR versus reduced radiation dose images reconstructed with MBIR. Discussion Fully iterative reconstruction such as MBIR better models complex reconstruction assumptions by utilizing advanced description of x-ray physics in the scanner and statistical noise in the images. Such assumptions include linear shape to the focal spot, fan character to the x-ray beam, and detectors that have both shape dimensions and variable response functions for the physics in the scanner (1,11,12). For the noise in the images, MBIR starts with back projection of estimated images from CT raw data, followed by forward projection of CT data from the estimated images. The forward projected data are then compared with the actual measured data with statistical metrics, and the computed difference is itself back projected to create an image update. This sequence is followed serially until the difference between actual measured data and new forward projected data decreases below an arbitrary threshold (1,5). The noise reduction capability of MBIR can be applied to low radiation dose CT data sets to produce images that are comparable in noise to standard radiation dose images reconstructed with ASIR. However, the different appearance of CT images reconstructed with MBIR may raise concerns about depiction of clinically relevant findings. The purpose of this study was to investigate whether reduced radiation dose liver CT images reconstructed with MBIR might compromise depiction of clinically relevant liver findings or compromise image quality when compared with standard dose CT images reconstructed with ASIR. We found that clinically reportable findings identified on the images from the standard dose ASIR series were all identified by two separate reviewers on the images from the reduced dose MBIR series. In 94% of the 136 liver findings, subjective conspicuity, subjective spatial resolution, and subjective assessment of image noise were assigned a score as the same or better with MBIR than with ASIR; none were assigned scores as nonevaluable. Subjective overall image quality and contour-definition and structure-delineation were assigned a score that was comparable between standard dose ASIR images and reduced dose MBIR images by both reviewers. While measured noise in fat was less and measured SNR was greater with reduced dose MBIR images compared with standard dose ASIR images, measured CNR was the same. The mean SSDE for the reduced radiation dose portal venous pass images reconstructed with MBIR was 59% lower than the standard radiation dose pass images reconstructed with ASIR (6.7 mgy vs 16.4 mgy). Our imaging protocol for both ASIR and MBIR involved arbitrary selection of some scanning parameters. First, the selection of 40% ASIR blended with FBP and a noise index of 41 HU for our routine clinical multiphase liver protocol was based on recommendations from the manufacturer and our own clinical experience during 3 years. Second, we thought it might be reasonable to target a 60% radiation dose reduction with MBIR reconstruction on the basis of abstracts presented at international meetings in 2010 and 2011, in addition to our own experience (13 15). Using previously published relationships between patient radiation dose and noise index values, we estimated that a noise index of 66 HU with MBIR might result in the targeted 60% patient dose reduction when averaged over multiple patients (16). Other authors have performed similar studies with MBIR in the abdomen. Deak et al (4) scanned the abdomen in 22 patients with a low radiation dose technique and also scanned a lowcontrast phantom, reconstructing images with FBP, ASIR, and MBIR. They found that MBIR markedly decreased image noise, improved subjective and objective image quality, and improved low contrast detectability, but it required minutes to reconstruct the images. Volders et al (17) recently reported scanning 51 patients with colorectal liver metastases by using a CT technique with a radiation dose reduced from their standard of care radiation dose by 2.36 mgy and reconstructed images with both 50% ASIR and MBIR. For lesions smaller than radiology.rsna.org n Radiology: Volume 273: Number 3 December 2014

7 Table 5 Readers B and C: Subjective Image Quality Scores for Standard Radiation Dose ASIR and Reduced Radiation Dose MBIR Score Standard Dose ASIR mm, conspicuity and detection were significantly greater with MBIR. For larger lesions, there was no difference. Chang et al (6) scanned patients who were suspected of having liver metastases (30 patients, by using a 50% reduced radiation dose protocol; and 70 patients, by using a standard dose protocol) in the portal venous phase. Images were reconstructed with MBIR, ASIR, and FBP. The reduced radiation dose MBIR images showed less noise, higher CNR, and better subjective image quality than reduced dose ASIR or FBP images and provided similar image quality compared with a standard dose FBP. Pickhardt et al (3) prospectively scanned the liver in 45 patients with a standard radiation dose image reconstructed with FBP and compared that reference scan with a follow-up subsequent CT scan obtained with a dose reduction of 57% 88% (mean, 74%) and reconstructed with FBP, 40% ASIR, and MBIR. Mean subjective and objective image quality and focal lesion detection were greatest for low radiation dose MIBR. Singh et al (18) scanned 10 patients with a double-pass technique during the portal venous phase: Images in the first pass were scanned at 200 mas and reconstructed with FBP, and images in the second pass were scanned at 50 mas and reconstructed with FBP, ASIR, and MBIR. They found Reduced Dose MBIR Reader B Reader C Reader B Reader C Overall image quality 1, nonevaluable , poor (2) 0 3, fair and sufficient 10 (20) 21 (42) 7 (14) 7 (14) 4, good 40 (80) 29 (58) 42 (84) 43 (86) Contour and delineation 1, severe blurring , moderate blurring , minimal blurring 3 (6) 15 (30) 12 (24) 15 (30) 4, no blurring 47 (94) 35 (70) 38 (76) 35 (70) Note. Numbers are frequencies of scores, with n = 50. Percentages of totals are in parentheses, and percentages were rounded. that the low radiation dose MBIR images subjectively and objectively resulted in acceptable image quality and in diagnostic confidence compared with the standard dose FBP images. As in our study, all of these other studies demonstrated that reduced dose MBIR resulted in better or acceptable image quality. This was true when we compared images from the same CT data set that were reconstructed with MBIR versus other algorithms. This was also true when we compared reconstruction of a reduced radiation dose CT data set with MBIR versus reconstruction of a standard radiation dose data set with other algorithms. This study had several important limitations. First, the number of patients is relatively small with a relatively narrow range of body mass index. A similar study in a greater number of patients that includes very small and very large subjects may produce different results. Second, we studied depiction of findings only in the liver in patients with cirrhosis. A study involving findings in other organs or other disease processes may also produce different results. Third, some of the CT examination and reconstruction parameters for both standard clinical ASIR images and for the research MBIR images were arbitrary, and a single value (eg, 40% ASIR with noise index 41 HU) comparing different parameter Table 6 Noise Measurements and SNR and CNR Calculations Factor Standard Dose ASIR Reduced Dose MBIR P Value Noise in fat ,.001 SNR ,.001 CNR Note. Data are means 6 standard deviations. values may yield other results. Fourth, while the two liver CT passes were in the same patient and temporally close together during the portal venous phase of the same CT examination, there was about a 7-second time difference between these two passes, which may have altered the appearance of images somewhat and placed the MBIR images at a relative disadvantage because the ASIR images were obtained at a time considered to be clinically optimal. Finally, we did not investigate sensitivity and specificity for liver findings with each reconstruction technique and at each radiation dose level, but rather chose to determine whether differences in reconstructed image appearance altered depiction of specific findings. In conclusion, on 59% reduced radiation dose portal venous liver CT images reconstructed with MBIR, independent reviewers were able to identify all findings depicted on the standard radiation dose portal venous CT images reconstructed with ASIR. Subjective image quality scores and objective measurements were similar or better with reduced radiation dose MBIR. This datum suggests that MBIR reconstruction may allow substantial radiation dose reduction in portal venous phase liver CT without compromising depiction of clinically relevant findings and with acceptable or improved image quality. Disclosures of Conflicts of Interest: W.P.S. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: disclosed no relevant relationships. Other relationships: institution received a clinical research grant from GE Healthcare for funding other projects. K.T.C. disclosed no relevant relationships. J.M.B. Activities related to the present article: disclosed Radiology: Volume 273: Number 3 December 2014 n radiology.rsna.org 799

8 no relevant relationships. Activities not related to the present article: institution received a research grant unrelated to this project from GE Healthcare. Other relationships: disclosed no relevant relationships. L.M.M. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: institution received research funding for a project not related to this article from GE Heathcare. Other relationships: disclosed no relevant relationships. E.C. disclosed no relevant relationships. K. M. Koprowicz disclosed no relevant relationships. K. M. Kanal disclosed no relevant relationships. References 1. Hsieh J. Computed tomography: principles, design, artifacts, and recent advances. 2nd ed. Bellingham, Wash: SPIE/Wiley, 2009; Vardhanabhuti V, Loader RJ, Mitchell GR, Riordan RD, Roobottom CA. Image quality assessment of standard- and low-dose chest CT using filtered back projection, adaptive statistical iterative reconstruction, and novel model-based iterative reconstruction algorithms. AJR Am J Roentgenol 2013;200(3): Pickhardt PJ, Lubner MG, Kim DH, et al. Abdominal CT with model-based iterative reconstruction (MBIR): initial results of a prospective trial comparing ultralow-dose with standard-dose imaging. AJR Am J Roentgenol 2012;199(6): Deák Z, Grimm JM, Treitl M, et al. Filtered back projection, adaptive statistical iterative reconstruction, and a model-based iterative reconstruction in abdominal CT: an experimental clinical study. Radiology 2013;266(1): Shuman WP, Green DE, Busey JM, et al. Model-based iterative reconstruction versus adaptive statistical iterative reconstruction and filtered back projection in liver 64- MDCT: focal lesion detection, lesion conspicuity, and image noise. AJR Am J Roentgenol 2013;200(5): Chang W, Lee JM, Lee K, et al. Assessment of a model-based, iterative reconstruction algorithm (MBIR) regarding image quality and dose reduction in liver computed tomography. Invest Radiol 2013;48(8): Katsura M, Matsuda I, Akahane M, et al. Model-based iterative reconstruction technique for radiation dose reduction in chest CT: comparison with the adaptive statistical iterative reconstruction technique. Eur Radiol 2012;22(8): Nelson RC, Feuerlein S, Boll DT. New iterative reconstruction techniques for cardiovascular computed tomography: how do they work, and what are the advantages and disadvantages? J Cardiovasc Comput Tomogr 2011;5(5): Katsura M, Sato J, Akahane M, et al. Comparison of pure and hybrid iterative reconstruction techniques with conventional filtered back projection: image quality assessment in the cervicothoracic region. Eur J Radiol 2013;82(2): AAPM Task Group 204. Size specific dose estimates (SSDE) in pediatric and adult CT examinations. reports/rpt_204.pdf. Published Accessed May Yu Z, Thibault JB, Bouman CA, Sauer KD, Hsieh J. Fast model-based x-ray CT reconstruction using spatially nonhomogeneous ICD optimization. IEEE Trans Image Process 2011;20(1): 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): Husarik DB, Marin D, Samei E, et al. Radiation dose reduction in abdominal computed tomography during the late hepatic arterial phase using a model-based iterative reconstruction algorithm: how low can we go? Invest Radiol 2012;47(8): Lin X, Tanaka I, Li J, Ueno E, Shen Y, Chen K. Dose reduction potential with advanced reconstruction algorithms: assessment of image noise and image quality in abdominal CT [abstr]. In: Radiological Society of North America Scientific Assembly and Annual Meeting Program. Oak Brook, Ill: Radiological Society of North America, 2011; O Neill S, McWilliams SR, O Neill F, et al. A quantitative comparison of model-based iterative reconstruction, adaptive statistical iterative reconstruction, and filtered back projection for abdominal CT [abstr]. In: Radiological Society of North America Scientific Assembly and Annual Meeting Program. Oak Brook, Ill: Radiological Society of North America, 2011; Kanal KM, Stewart BK, Kolokythas O, Shuman WP. Impact of operator-selected image noise index and reconstruction slice thickness on patient radiation dose in 64-MDCT. AJR Am J Roentgenol 2007;189(1): Volders D, Bols A, Haspeslagh M, Coenegrachts K. Model-based iterative reconstruction and adaptive statistical iterative reconstruction techniques in abdominal CT: comparison of image quality in the detection of colorectal liver metastases. Radiology 2013;269(2): Singh S, Kalra MK, Do S, et al. Comparison of hybrid and pure iterative reconstruction techniques with conventional filtered back projection: dose reduction potential in the abdomen. J Comput Assist Tomogr 2012;36(3): radiology.rsna.org n Radiology: Volume 273: Number 3 December 2014