Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
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 www.rsna.org/rsnarights. ORIGINAL RESEARCH Sarabjeet Singh, MBBS, MMST Mannudeep K. Kalra, MD Jiang Hsieh, PhD Paul E. Licato, MS Synho Do, PhD Homer H. Pien, PhD Michael A. Blake, MD Purpose: To compare image quality and lesion conspicuity on abdominal computed tomographic (CT) images acquired with different x-ray tube current–time products (50–200 mAs) and reconstructed with adaptive statistical iterative reconstruction (ASIR) and filtered back projection (FBP) techniques. Materials and Methods: Twenty-two patients (mean age, 60.1 years 6 7.3 [standard deviation]; age range, 52.8–67.4 years; mean weight, 78.9 kg 6 18.3; 12 men, 10 women) gave informed consent for this prospective institutional review board–approved and HIPAA-compliant study, which involved the acquisition of four additional image series at multidetector CT. Images were acquired at different tube current–time products (200, 150, 100, and 50 mAs) and encompassed an abdominal lesion over a 10-cm scan length. Images were reconstructed separately with FBP and with three levels of ASIR-FBP blending. Two radiologists reviewed FBP and ASIR images for image quality in a blinded and randomized manner. Volume CT dose index (CTDIvol), dose-length product, patient weight, objective noise, and CT numbers were recorded. Data were analyzed by using analysis of variance and the Wilcoxon signed rank test. Results: CTDIvol values were 16.8, 12.6, 8.4, and 4.2 mGy for 200, 150, 100, and 50 mAs, respectively (P , .001). Subjective noise was graded as below average at 150 mAs and average at 100 and 50 mAs for ASIR images, as compared with FBP images, on which noise was graded as average at 150 mAs, above average at 100 mAs, and unacceptable at 50 mAs. A substantial blotchy image appearance was noted in four of 22 image series acquired at 4.2 mGy with 70% ASIR. Lesion conspicuity was significantly better at 4.2 mGy on ASIR than on FBP images (observed P , .044), and overall diagnostic confidence changed from unacceptable on FBP to acceptable on ASIR images. Conclusion: ASIR lowers noise and improves diagnostic confidence in and conspicuity of subtle abdominal lesions at 8.4 mGy when images are reconstructed with 30% ASIR blending and at 4.2 mGy in patients weighing 90 kg or less when images are reconstructed with 50% or 70% ASIR blending. 1 From the Department of Radiology, Massachusetts General Hospital, 25 New Chardon St, Suite 400B, Boston, MA 02114 (S.S., M.K.K., S.D., H.H.P., M.A.B.); and GE Healthcare, Waukesha, Wis (J.H., P.E.L.). Received December 7, 2009; revision requested January 12, 2010; revision received February 1; accepted February 10; final version accepted June 7. Address correspondence to S.S. (e-mail: [email protected] ). q q RSNA, 2010 Supplemental material: http://radiology.rsna.org/lookup /suppl/doi:10.1148/radiol.10092212/-/DC1 RSNA, 2010 Radiology: Volume 257: Number 2—November 2010 n radiology.rsna.org 373 n GASTROINTESTINAL IMAGING Abdominal CT: Comparison of Adaptive Statistical Iterative and Filtered Back Projection Reconstruction Techniques1 GASTROINTESTINAL IMAGING: Reconstruction Techniques at Abdominal CT I mage reconstruction algorithms play a critical role in the quality and appearance of tomographic images (1–4). Although iterative image reconstruction algorithms were used to generate images with the very first commercial clinical computed tomographic (CT) scanner (4) and underwent substantial improvements in the 1980s, especially in the context of emission tomography (1–3), filtered back projection (FBP) algorithms were used for CT image reconstruction owing to their faster image reconstruction and ease of implementation (5). Over the past decade, the desire for finer resolution, greater volume coverage, and faster scan times and the desire to concurrently lower radiation dose have pushed the performance of FBP reconstruction to its limits. Alternative techniques of image reconstruction with iterative reconstruction algorithms have recently become available for clinical use. In contrast to FBP, iterative reconstruction uses a forward reconstruction model and a more precise modeling of scanner geometry and the underlying physics. Results of prior studies (6–13) have shown that iterative reconstructions produce higher-resolution images with greater robustness for various imaging artifacts with a longer computational processing time than FBP reconstruction. The increase in reconstruction time Advances in Knowledge n Abdominal CT images acquired at 100–200 mAs and reconstructed with adaptive statistical iterative reconstruction (ASIR) have acceptable image noise and enable acceptable diagnostic confidence. n Higher blending factors of 70% ASIR and 30% filtered back projection (FBP) are required to obtain acceptable image noise and diagnostic confidence at lower radiation doses. n Abdominal CT images tended to have a blotchy pixilated appearance at the higher blending proportion of 70% ASIR and 30% FBP. 374 Singh et al and the need for costlier computation hardware presently makes iterative reconstructions unacceptable for clinical use. To circumvent the issue of the prolonged reconstruction time and the need for higher computational power, the adaptive statistical iterative reconstruction (ASIR) algorithm has been developed. This algorithm takes into account precise modeling of the x-ray photon statistics and electronic noise, all of which are less accurate in FBP (Appendix E1 [online]). To decrease reconstruction time, ASIR utilizes the information contained in the FBP-reconstructed image as an initial “building block” in the reconstruction process (14). Reduction of noise and artifacts in low-radiationdose images are especially helpful in young and obese patients, in clinical situations like multiphase CT examinations, and in patients who have undergone multiple prior examinations. The ASIR technique has been approved by the U.S. Food and Drug Administration for clinical use. The purpose of our study was to compare image quality, lesion detection, and lesion conspicuity on abdominal CT images acquired at different x-ray tube current–time products (50–200 mAs) and reconstructed with ASIR and FBP. Materials and Methods M.K.K. received research funding from GE Healthcare (Waukesha, Wis). J.H. and P.E.L. are employees of GE Healthcare. The remaining authors (S.S., S.D., H.H.P., and M.A.B.) have no pertinent disclosures and had full control of the study data during the study. Implications for Patient Care n Reduction of CT radiation dose down to 8.4 mGy is feasible for abdominal CT images reconstructed with ASIR without compromising image quality, lesion detection, and lesion conspicuity; for patients weighing 90 kg or less, radiation dose reduction down to 4.2 mGy is possible with the ASIR technique. Patients This prospective clinical study was approved by the Human Research Committee of our institutional review board and was conducted in compliance with Health Insurance Portability and Accountability Act guidelines. Written informed consent was obtained from all 22 patients (mean age, 60.1 years 6 7.3; age range, 52.8–67.4 years; mean weight, 78.9 kg 6 18.3; 12 men [mean age, 60.9 years 6 7] and 10 women [mean age, 59.3 years 6 8]) for the acquisition of four extra sets of images, in addition to their clinically indicated standard-of-care abdominal CT examinations. All consecutive eligible patients who gave informed consent for participation in the study were included in the study, which was performed between February 2, 2009 and August 4, 2009. Only those patients who met all of the following inclusion criteria for the study were recruited: age greater than 50 years, scheduled for CT of the abdomen, able to provide informed written consent, able to hold their breath for at least 10 seconds, able to follow verbal commands for breath holding and remain still for the duration of scanning, and hemodynamic stability (conscious and oriented, with a regular respiration rate of 12–40 breaths per minute, a pulse rate of 60–90 beats per minute Published online before print 10.1148/radiol.10092212 Radiology 2010; 257:373–383 Abbreviations: ASIR = adaptive statistical iterative reconstruction CTDIvol = volume CT dose index FBP = filtered back projection MTF = modulation transfer function Author contributions: Guarantor of integrity of entire study, M.K.K.; study concepts/study design or data acquisition or data analysis/ interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, S.S., M.K.K., J.H., H.H.P., M.A.B.; clinical studies, S.S., M.K.K., M.A.B.; experimental studies, S.S., M.K.K., S.D., H.H.P.; statistical analysis, S.S., M.K.K., H.H.P.; and manuscript editing, S.S., M.K.K., J.H., P.E.L., H.H.P., M.A.B. See Materials and Methods for pertinent disclosures. radiology.rsna.org n Radiology: Volume 257: Number 2—November 2010 GASTROINTESTINAL IMAGING: Reconstruction Techniques at Abdominal CT without dysrhythmia, and a systolic blood pressure of 100–140 mm Hg). To avoid variations in the postcontrast timing of various abdominal CT protocols and to minimize radiation dose, only patients who were undergoing single-phase abdominal CT with image acquisition were included in our study. We excluded patients from our study if they were younger than 50 years of age, were hemodynamically unstable, or were not scheduled for routine CT of the abdomen. Patients were also excluded if they were unable to provide informed written consent, follow verbal commands for breath holding, or remain still for the duration of scanning. To recruit patients for the study, we checked our radiology information system to identify patients scheduled for standard-of-care clinical CT examinations with the scanner used in our study. Ten to 15 such patients were identified per day. We then applied our inclusion and exclusion criteria to find eligible patients. Most of the patients included in our study underwent abdominal CT scanning for clinical indications that included staging of known malignancy (primary sarcoma, breast, colon, prostate carcinoma) or suspected malignancy (chest or abdominal), renal stones, change of bowel habits, and ruling out abdominal pain. Next, we requested permission from the physicians of the eligible patients (five or six per day) for participation in the study protocol. Of these eligible patients, only 25 gave their informed consent for participation in the study. Only one patient refused to undergo this research study after signing the consent form but prior to the acquisition of research images. Therefore, no research CT images were acquired in this patient. Thus, our final study group, in whom qualitative and quantitative image quality was assessed for comparison of the ASIR and FBP techniques, consisted of 22 patients. Scanning Techniques First, a clinically indicated standardof-care abdominal CT examination was performed with a commercial CT scanner (Discovery CT750 HD; GE Healthcare) and administration of 80–100 mL Radiology: Volume 257: Number 2—November 2010 n of an intravenous contrast medium (Iopamidol 370; Bracco Diagnostics, Princeton, NJ). Subsequently, four additional sets of images were acquired in each patient for the purpose of this study. Each of the additional image data sets was acquired through an identical part of the abdomen over a scan length of 10 cm. The location of the acquisition of these additional image data sets was based on a review (M.K.K., a radiologist with 8 years of experience) of the patient’s prior and current standard-ofcare abdominal CT images to include the most subtle or smallest lesion in the abdomen. All patients were scanned at the level of either the lower liver (at or below the porta hepatis) or the kidneys, as all the assessed lesions of interest were located at these levels. None of the image series was acquired through the lung bases. As per the standard of care, scanning in the portal venous phase was triggered by using a bolus-tracking technique. The time between completion of the standard-of-care abdominal CT examination and the initiation of scanning for research image acquisition was 10 or fewer seconds in most patients. All research images were acquired within 1 minute of the standardof-care abdominal CT examination. The interval between the acquisitions of the four research image series was 6–10 seconds. Although these measures minimized the time between acquisitions of research images and standard-of-care dynamic contrast material–enhanced CT images, most research images were acquired in an early equilibrium phase, as it was not possible to perform focused dynamic image acquisition at four different dose levels. To compare the FBP and ASIR techniques, abdominal CT images were acquired at four radiation dose levels by selecting four fixed tube current–time products (200, 150, 100, and 50 mAs). Although automatic exposure control techniques are routinely used for dose modulation in typical clinical practice, we did not use these techniques to achieve radiation dose reductions in our study for several reasons. First, to obtain four levels of radiation dose with automatic exposure control with the CT radiology.rsna.org Singh et al scanner used in our study, we would have been required to adjust the noise index (a descriptor for desired image noise used on the scanner to adjust tube current), minimum tube current, and maximum tube current for each dose level. Second, these factors would have had to be tailored to individual patient size because at a constant noise index, automatic exposure control increases the radiation dose to larger patients and decreases the dose to smaller patients. Furthermore, this increase or decrease in radiation dose with automatic exposure control is intentionally kept as nonlinear in the actual CT scanners to avoid the inadvertent use of high tube current in larger patients (15) and, paradoxically, to prevent the use of too low a tube current in smaller patients. Also, results of prior studies (10) have indeed shown that other iterative reconstruction techniques can maintain a similar noise level regardless of radiation dose levels at up to a ninefold reduction in dose. Compared with the use of a fixed tube current, it is more difficult to control the extent of dose reduction with an automatic exposure control technique with adjustment in the desired image quality metric such as noise index (GE Healthcare) or reference tube current–time product. To avoid contrast enhancement bias due to the delay in scanning from the start of injection, the acquisition sequence of the four research CT data sets was randomized. Because it was not possible or practical to inject contrast medium four times in a patient to acquire images at identical contrast enhancement phases, no additional intravenous contrast medium was administered for acquisition of the research image series. All scanning parameters, with the exception of tube current, were held constant and included 120 kVp, a pitch of 0.984:1, a 39.37-mm table speed per gantry rotation, a helical acquisition mode, a detector configuration of 64 detector rows with 0.625-mm-thick sections, a 0.5-second gantry rotation time, a reconstructed section thickness of 5 mm, a reconstruction section interval of 5 mm, and a standard reconstruction kernel. 375 GASTROINTESTINAL IMAGING: Reconstruction Techniques at Abdominal CT Singh et al Figure 1 Figure 1: Flowchart shows CT image reconstruction in different radiation dose levels and the varying levels of FBP and ASIR blending (ASIR 30, ASIR 50, and ASIR 70) for each radiation dose level. Image Reconstruction Each of the four raw research CT data sets was reconstructed by using FBP to generate four FBP image series (one each at four tube current–time products: 200, 150, 100, and 50 mAs) (Fig 1). The four research raw scan data sets were also reconstructed by using the ASIR technique. ASIR image reconstruction was selected for three different levels of blending (30%, 50%, and 70%) for each of four radiation dose levels to obtain 12 sets of research ASIR image data sets (30% ASIR at 200 mAs, 50% ASIR at 200 mAs, 70% ASIR at 200 mAs, 30% ASIR at 150 mAs, 50% ASIR at 150 mAs, 70% ASIR at 150 mAs, 30% ASIR at 100 mAs, 50% ASIR at 100 mAs, 70% ASIR at 100 mAs, 30% ASIR at 50 mAs, 50% ASIR at 50 mAs, and 70% ASIR at 50 mAs). Thus, 352 image series (22 patients times 16 image data sets) were generated. There was a delay of about 5 seconds in the processing and display of the first ASIR image in the series, but the frame rate of display of the ASIR images was quite similar to that of the FBP images, at up to 12 frames per second. To enable double-blinded evaluation, each image data set was coded, deidentified, and randomized by a study coauthor (S.S., with 3 years of experience). The blending of ASIR and FBP techniques was preferred to pure ASIR, as 376 the ASIR technique alone causes substantial changes in noise texture, which was found to be undesirable by a group of radiologists in an advisory board meeting of the vendor (GE Healthcare). Blending of the two reconstruction techniques helped preserve the image texture, or the “look,” of FBP images, while the use of ASIR reduced image noise. As higher blending proportions of ASIR to FBP (ie, .70%) would have substantially changed the texture and characteristics of the images, on the basis of the vendor’s recommendation, we limited assessment of image quality to 70% ASIR. Subjective Image Quality All randomized CT image data sets were reviewed at a picture archiving and communication systems diagnostic workstation (Impax ES; Agfa Technical Imaging Systems, Ridgefield Park, NJ) for assessment of subjective image quality. All image data sets were presented in blinded and randomized manner to two experienced radiologists (M.A.B. [with 12 years of experience] and M.K.K.) for independent assessment of image quality. Both radiologists were trained on two image data sets for the grading of different aspects of subjective image quality and lesion assessment so that they would understand the evaluation system, as well as to improve interobserver agreement. These two image data sets belonged to the first two patients recruited in our study and were not used for the subsequent statistical analyses. Both radiologists assessed these two image data sets just before the evaluation of the rest of the CT studies that were used in the statistical analyses. Subjective image quality was assessed in terms of subjective image noise by using a five-point scale (1 = minimal image noise, 2 = less than average noise, 3 = average image noise, 4 = above average noise, and 5 = unacceptable image noise). Each artifact was graded by using a four-point scale (1 = no artifacts, 2 = minor artifacts not interfering with diagnostic decision making, 3 = major artifacts affecting visualization of major structures but diagnosis still possible, and 4 = artifacts affecting diagnostic information). The following artifacts were assessed: helical or windmill artifacts; streak artifacts due to metals and leads; beam hardening artifacts due to the patient having his arms by his side; truncation due to large body size or off centering (this was rare); and a blotchy, pixilated appearance. The visibility of small structures (small blood vessels, adrenal glands, small lymph nodes) was ranked by using a five-point scale (1 = excellent visualization, 2 = above average visibility, 3 = acceptable visibility, 4 = suboptimal visibility, and 5 = unacceptable visualization), while lesion size was measured by using a four-point scale (1 = focal and , 1 cm, 2 = focal and 1–5 cm, 3 = focal and . 5 cm, and 4 = diffuse lesion), Subjective visual lesion conspicuity was assessed by using a fivepoint scale (1 = well-seen lesion with well-visualized margins, 2 = well-seen lesion with poorly visualized margins, 3 = subtle lesion, 4 = probably an artifact mimicking a lesion, and 5 = definitely an artifact mimicking a lesion), image contrast was assessed by using a five-point scale (1= excellent image contrast, 2 = above average contrast, 3 = acceptable image contrast, 4 = suboptimal image contrast, and 5 = very poor contrast), and diagnostic confidence was assessed by using a four-point scale (1 = completely confident; 2 = probably confident; 3 = confident only for a limited clinical entity such as a kidney stone, a calcified lesion, radiology.rsna.org n Radiology: Volume 257: Number 2—November 2010 GASTROINTESTINAL IMAGING: Reconstruction Techniques at Abdominal CT Singh et al Table 1 Percentage Agreement between the Two Radiologists for Each of the Assessed Subjective Image Quality and Lesion Assessment Parameters Reconstruction Technique and Tube Current–Time Product (mAs) FBP 200 150 100 50 30% ASIR 200 150 100 50 50% ASIR 200 150 100 50 70% ASIR 200 150 100 50 Confidence in Diagnosis Image Contrast Lesion Conspicuity Noise Visibility of Small Structures 95.4 (21/22) 100 (22/22) 95.4 (21/22) 86.4 (19/22) 100 (22/22) 100 (22/22) 100 (22/22) 72.2 (16/22) 95.4 (21/22) 90.9 (20/22) 81.8 (18/22) 81.8 (18/22) 100 (22/22) 90.9 (20/22) 72.2 (16/22) 95.4 (21/22) 100 (22/22) 100 (22/22) 100 (22/22) 100 (22/22) 100 (22/22) 100 (22/22) 100 (22/22) 86.4 (19/22) 95.4 (21/22) 95.4 (21/22) 95.4 (21/22) 72.2 (16/22) 100 (22/22) 90.9 (20/22) 81.8 (18/22) 77.2 (17/22) 100 (22/22) 95.4 (21/22) 86.4 (19/22) 95.4 (21/22) 100 (22/22) 100 (22/22) 100 (22/22) 86.4 (19/22) 100 (22/22) 100 (22/22) 100 (22/22) 86.4 (19/22) 95.4 (21/22) 95.4 (21/22) 95.4 (21/22) 77.2 (17/22) 100 (22/22) 90.9 (20/22) 81.8 (18/22) 77.2 (17/22) 100 (22/22) 100 (22/22) 100 (22/22) 81.8 (18/22) 100 (22/22) 100 (22/22) 100 (22/22) 77.2 (17/22) 100 (22/22) 100 (22/22) 100 (22/22) 86.4 (19/22) 95.4 (21/22) 95.4 (21/22) 100 (22/22) 72.2 (16/22) 100 (22/22) 90.9 (20/22) 81.8 (18/22) 77.2 (17/22) 100 (22/22) 100 (22/22) 100 (22/22) 100 (22/22) 95.4 (21/22) 95.4 (21/22) 95.4 (21/22) 77.2 (17/22) Note.—Raw data are in parentheses. or a large lesion; and 4 = poor confidence). The image quality attributes assessed in our study have been described in the European Guidelines on Quality Criteria for Computerized Tomography document (16) and have been used in multiple prior studies in the radiology literature (17). To avoid bias in lesion detection at four dose levels in 16 image series, radiologists were first asked to grade the subjective image noise for individual image series and then to assess lesion detection, beginning with the image series with the highest image noise. Neither radiologist knew about the presence of lesions in the patients. A standardized template for lesion assessment could not be used, as different patients had different body regions imaged and images were acquired through limited scan lengths. Objective Measurements Objective image noise (in Hounsfield units) 6 standard deviations and CT numbers (Hounsfield units) were measured for 352 CT image series. Circular Radiology: Volume 257: Number 2—November 2010 n regions of interest (20–30 mm) were drawn in the abdominal aorta, without touching the lumen walls, to cover at least two-thirds of its lumen. Circular regions of interest (20–30 mm) were also drawn in the homogeneous area of the right lobe of the liver. The skin-to-skin maximum transverse diameter of the abdomen was measured from localizer radiographs, as transverse images are often reconstructed with a smaller field of view and may not include the skin. Each patient was weighed by using a digital weighing machine just prior to the CT examination. CT radiation dose descriptors such as volume CT dose index (CTDIvol, described in milligrays) and doselength product (described in mGy · cm) were recorded after completion of the CT examination for all image data sets. Estimation of Modulation Transfer Function To compare the effect of ASIR and FBP reconstructions on spatial resolution, we performed a phantom study to estimate the modulation transfer function (MTF) radiology.rsna.org by using the same scanner used for imaging patients that is described above. A phantom (Catphan600; The Phantom Laboratory, Salem, NY) with 28-mmdiameter tungsten wire was scanned at a tube current of 120 kVp, with 200 mA, a helical acquisition mode, a section thickness of 0.625 mm, a 0.5-second gantry rotation time, and a standard reconstruction algorithm. Images were reconstructed with the FBP technique and with 30%, 50%, and 70% ASIR techniques. The MTF was measured as the angular average of the two-dimensional Fourier transform of the point spread function in each of the reconstructed image data sets. Statistical Analysis Data were analyzed by using analysis of variance for objective metrics such as objective image noise and CT numbers and by using the Wilcoxon signed rank test for subjective image quality and lesion assessment parameters. Intraobserver variability was not estimated, as each radiologist assessed 377 GASTROINTESTINAL IMAGING: Reconstruction Techniques at Abdominal CT Singh et al Table 2 Subjective Image Quality Scores and Frequency of Scores for Image Noise, Lesion Conspicuity, and Diagnostic Confidence Reconstruction Technique and Tube Current–Time Product (mAs) FBP 200 150 100 50 30% ASIR 200 150 100 50 50% ASIR 200 150 100 50 70% ASIR 200 150 100 50 Image Noise Lesion Conspicuity Diagnostic Confidence 2 (16/22), 3 (6/22) 3 (16/22), 4 (3/22), 5 (3/22) 4 (16/22), 3 (3/22), 5 (3/22) 5 (19/22), 4 (2/22), 3 (1/22) 1 (18/22), 3 (3/22), 2 (1/22) 1 (17/22), 3 (4/22), 2 (1/22) 1 (15/22), 3 (4/22), 2 (3/22) 3 (14/22), 1 (6/22), 2 (1/22), 4 (1/22) 1 (22/22) 1 (20/22), 2 (1/22), 3 (1/22) 2 (17/22), 1 (4/22), 3 (1/22) 4 (18/22), 1 (3/22), 2 (1/22) 2 (19/22), 3 (2/22), 1 (1/22) 3 (14/22), 2 (5/22), 4 (3/22) 3 (15/22), 2 (1/22), 4 (6/22) 4 (15/22), 5 (5/22), 3 (2/22) 1 (18/22), 3 (3/22), 2 (1/22) 1 (17/22), 3 (4/22), 2 (1/22) 1 (16/22), 3 (4/22), 2 (2/22) 1 (13/22), 3 (5/22), 2 (3/22), 4 (1/22) 1 (22/22) 1 (22/22) 2 (16/22),1 (6/22) 2 (18/22), 1 (2/22), 4 (1/22), 3 (1/22) 1 (19/22), 2 (3/22) 2 (18/22), 3 (2/22), 1 (2/22) 2 (16/22), 3 (4/22), 1 (1/22), 4 (1/22) 4 (16/22), 3 (4/22), 2 (2/22) 1 (17/22), 3 (3/22), 2 (2/22) 1 (17/22), 3 (4/22), 2 (1/22) 1 (16/22), 3 (4/22), 2 (2/22) 1 (14/22), 3 (5/22), 2 (2/22), 4 (1/22) 1 (22/22) 1 (20/22), 2 (2/22) 1 (20/22), 2 (2/22) 2 (14/22), 1 (4/22), 3 (3/22), 4 (1/22) 1 (22/22) 1 (19/22), 3 (2/22), 2 (1/22) 2 (17/22), 1 (4/22), 3 (1/22) 3 (16/22), 2 (3/22), 1 (2/22), 4 (1/22) 1 (17/22), 3 (3/22), 2 (2/22) 1 (17/22), 3 (4/22), 2 (1/22) 1 (16/22), 3 (4/22), 2 (2/22) 1 (15/22), 3 (5/22), 2 (1/22), 4 (1/22) 1 (21/22), 2 (1/22) 1 (19/22), 2 (3/22) 1 (19/22), 2 (3/22) 1 (13/22), 2 (4/22), 3 (3/22), 4 (2/22) Note.—Data are scores, with the frequency of each score in parentheses. the images only once. Interobserver variability was estimated by using both k statistics and percentage agreement between the two radiologists for each of the assessed subjective image quality and lesion assessment parameters. Definitions of levels of agreement on the basis of k values were as follows: k , 0.20 indicated slight agreement; k = 0.20–0.40, fair agreement; k = 0.41– 0.60, moderate agreement; k = 0.61– 0.80, substantial agreement; and k = 0.81–1.0, almost perfect agreement. To determine the effect of patient size on subjective image quality and objective image noise on ASIR and FBP images, we arbitrarily classified patients into two groups (those weighing ⱕ 90 kg and those weighing . 90 kg). P , .05 was considered to indicate a statistically significant difference. Results There was no significant difference in size between the 12 male patients (mean age, 59.5 years 6 6.8; mean weight, 90.3 kg 6 17.5; and mean transverse 378 diameter, 43.1 cm 6 4.8) and the 10 female patients (mean age, 65.9 years 6 7.3; mean weight, 71.7 kg 6 11.5; and mean transverse diameter, 41.3 cm 6 3.8) (observed P = .3). There was variable interobserver agreement between the two radiologists (k = 0.12–1.0). However, as summarized in Table 1, the percentage agreement between the two radiologists ranged from 72.2% (16 of 22 scores in agreement for visibility of small structures) to 100% (22 of 22 scores in agreement for criteria such as image noise, lesion conspicuity, and diagnostic confidence). Subjective Image Quality Detailed image quality and lesion detection scores are summarized in Table 2. Both radiologists ranked subjective image noise as suboptimal or unacceptable in FBP images obtained at 50 and 100 mAs (k = 0.12 at 50 mAs), while noise was ranked as average or acceptable in 30% ASIR images at 100–200 mAs (k = 0.59 at 100 mAs; k = 0.8 at 150 mAs), in 50% ASIR images ( k = 1.0 at 100 mAs), and 70% ASIR images. Only 70% ASIR images were rated as acceptable for image noise at 50 mAs. No major artifacts were seen in any of the ASIR or FBP image series acquired at the four dose levels. Minor beam hardening or photon starvation artifacts were noted in both ASIR and FBP images in three of 22 patients (mean weight, 97.1 kg 6 23.4). A minor, blotchy pixilated appearance of the images was seen in two CT studies at 50 and 100 mAs with 30% ASIR and in three CT studies at 50–150 mAs with 50% ASIR (Table 3). No such appearance was seen at 200 mAs. On the other hand, a blotchy, pixilated appearance was substantial in four of the 22 CT image series acquired at 100 and 50 mAs and reconstructed by using 70% ASIR. In three of four image series, this appearance did not affect diagnostic interpretation, while in one series acquired in a patient of average size (weight = 77 kg), it interfered with the diagnostic confidence of both radiologists (Fig 2). Visibility of small abdominal structures was rated as acceptable at all doses with both FBP radiology.rsna.org n Radiology: Volume 257: Number 2—November 2010 GASTROINTESTINAL IMAGING: Reconstruction Techniques at Abdominal CT Figure 2 Figure 3 Figure 2: Transverse abdominal CT images. (a) Image reconstructed with 50% ASIR is diagnostically acceptable, but (b) image reconstructed with 70% ASIR shows a blotchy, pixilated image texture and irregular edges, rendering it diagnostically unacceptable. Figure 3: Transverse abdominal CT images in 51-year-old woman weighing 63 kg with a right hepatic lobe enhancing lesion (hemangioma) reconstructed with FBP and three levels of ASIR (30%, 50%, and 70%) at four tube current–time products (200, 150, 100, and 50 mAs). Image noise and artifacts were reduced with ASIR, and images were diagnostically acceptable even at 50 mAs with 70% ASIR. Table 3 Scores for Blotchy, Pixilated Imaging Appearance of 22 Abdominal CT Studies Reconstruction Technique and ASIR (k = 0.62 for 30% ASIR at 50 mAs; k = 0.29 for 70% ASIR at 50 mAs). Image contrast was found to be acceptable at all radiation dose levels and ASIR blending levels (k = 0.24 for FBP at 50 mAs; k = 0.24 for 30% ASIR at 50 mAs; k = 0.3 for 50% ASIR at 50 mAs; k = 0.63 for 70% ASIR at 200 mAs; k = 0.23 for 70% ASIR at 50 mAs) (Figs 3 and 4). No lesions were missed on FBP or ASIR images. Detected lesions in our study included subcentimeter focal renal cysts and masses (n = 23 lesions), adrenal lesions (n = 6), vertebral lesions (n = 6), focal liver lesions (n = 5), abdominal lymph nodes (n = 3), parapelvic cysts (n = 2), gallbladder stones (n = 1), diverticulosis (n = 1), and a dilated main pancreatic duct (n = 1). Of the 68 abdominal lesions detected in 22 patients, 48 were less Radiology: Volume 257: Number 2—November 2010 Singh et al n FBP 30% ASIR 50% ASIR 70% ASIR 200 mAs 150 mAs 100 mAs 50 mAs 22/0/0/0 22/0/0/0 22/0/0/0 22/0/0/0 22/0/0/0 21/1/0/0 19/3/0/0 7/14/1/0 22/0/0/0 20/2/0/0 19/3/0/0 7/11/3/1 22/0/0/0 20/2/0/0 19/3/0/0 7/11/3/1 Note.—Data are numbers of CT studies given pixilation scores of 1, 2, 3, and 4, respectively. Because discrete values cannot be averaged, the lowest or similar scores for pixilation given by the radiologists are presented. than 1 cm in size, and two were between 1 and 5 cm. Of the 68 lesions, 65 had well-visualized margins, while the remaining three were subtle, low-contrast renal lesions. Lesions with conspicuity that was graded as well seen with wellvisualized margins on ASIR images at 50–200 mAs (k = 0.41 for 30% ASIR at 50 mAs; k = 0.25 for 30% ASIR at 100 mAs; k = 0.44 for 30% ASIR at 150 mAs; k = 0.12 for 50% ASIR at 50 mAs; k = 0.25 for 50% ASIR at 100 mAs; k = 0.44 for 50% ASIR at 150 mAs; k = 0.11 for 70% ASIR at 50 mAs; k = 0.25 radiology.rsna.org for 70% ASIR at 100 mAs; k = 0.44 for 70% ASIR at 150 mAs) were graded as well seen with well-visualized margins on FBP images at 100–200 mAs (k = 0.25 at 100 mAs; k = 0.44 at 150 mAs) but as subtle in FBP images at 50 mAs (k = 0.25) (observed P , .044). Diagnostic confidence was unacceptable on FBP images at 50 mAs (k = 0.29), whereas it improved to a grade of “probably acceptable” on 30% ASIR images (k = 0.29) and 50% ASIR images (k = 0.29) and “fully acceptable” on 70% ASIR images (k = 0.29). 379 GASTROINTESTINAL IMAGING: Reconstruction Techniques at Abdominal CT Singh et al Table 4 Objective Image Noise in the Liver in 22 Abdominal CT Studies Reconstruction Technique FBP 30% ASIR 50% ASIR 70% ASIR 200 mAs 150 mAs 100 mAs 50 mAs 14.5 6 0.7 (13.8–15.2) 11.3 6 0.6 (10.6–12.0) 9.3 6 0.5 (8.8–9.8) 7.2 6 0.4 (6.8–7.7) 18.7 6 0.8 (17.9–19.5) 14.9 6 0.6 (14.3–15.5) 12.2 6 0.5 (11.7–12.7) 9.5 6 0.4 (9.1–9.9) 23.8 6 0.9 (22.9–24.7) 18.3 6 0.7 (17.5–19.0) 15 6 0.6 (14.4–15.6) 11.7 6 0.6 (11.1–12.3) 32.4 6 1.2 (31.1–33.6) 25.8 6 1.0 (24.8–26.8) 20.7 6 0.9 (19.8–21.6) 16.2 6 0.8 (15.2–16.8) Note.—Data are mean objective image noise (in Hounsfield units) 6 standard deviation, with ranges in parentheses. Objective image noise decreased in the liver with an increase in ASIR level of image reconstruction at all four tube current–time products. Objective Image Quality Detailed objective image quality values are summarized in Tables 4 and 5. At 200 mAs, mean objective image noise 6 standard error of the mean decreased by about 50.3% in liver (to 7.2 6 0.4 from 14.5 6 0.7) and by about 53.5% (to 8.5 6 0.4 from 18.3 6 0.8) in the aorta for 70% ASIR images as compared with FBP images (P , .001). At 50 mAs, image noise decreased by about 50% (to 16.2 6 0.8 from 32.4 6 1.2) in the liver and by about 52.2% (to 18.9 6 0.7 from 39.6 6 1.3) in the aorta for 70% ASIR images as compared with FBP images (P , .001). Contrary to FBP images, ASIR images had less objective and subjective image noise, regardless of patient weight (observed P , .001 for both comparisons) (Table 6). Subjective image noise was deemed unacceptable for patients weighing greater than 90 kg for FBP images at 50 mAs and 30% ASIR images. Although image noise was rated as acceptable for all patient sizes at all tube current– time products with 50% and 70% ASIR blending, 50% ASIR images at 50 mAs were not fully acceptable in terms of diagnostic confidence for patients heavier than 90 kg, and greater pixilation was noted in 70% ASIR images at 50 mAs. There was no significant change in the average CT numbers regardless of the tube current–time product (200–50 mAs) and reconstruction technique (FBP, 30% ASIR, 50% ASIR, and 70% ASIR) (observed P , .323). improvement in the spatial resolution of images reconstructed with ASIR at all levels compared with the FBP images at both 10% and 50% MTF. The spatial resolution improved slightly with increasing strength of ASIR from 30% to 70%. MTF Estimation Results of the estimation of the MTF, or spatial resolution, are summarized in Figure 5. There was small but consistent Radiation Doses CTDIvol values at 200, 150, 100, and 50 mAs were 16.8, 12.6, 8.4, and 4.2 mGy, respectively (P , .001). Dose-length pro- 380 Figure 4 Figure 4: Transverse abdominal CT images in 65-year-old man weighing 70 kg with multiple hypointense renal lesion (cysts) reconstructed with FBP and three levels of ASIR (30%, 50%, and 70%) at four tube current–time products (200, 150, 100, and 50 mAs). Even images acquired at 50 mAs and reconstructed with 70% ASIR had lower image noise and artifacts than FBP images at the same tube current–time product and radiation dose level. ducts for 200, 150, 100, and 50 mAs were 237.1, 177.8, 118.5, and 59.3 mGy · cm, respectively (P , .001). Discussion Owing to radiation dose concerns associated with CT, several efforts have been made to reduce radiation dose without compromising the quality of diagnostic information; these efforts include lowering the tube current–time product radiology.rsna.org n Radiology: Volume 257: Number 2—November 2010 GASTROINTESTINAL IMAGING: Reconstruction Techniques at Abdominal CT Singh et al Table 5 Objective Image Noise in the Aorta in 22 Abdominal CT Studies Reconstruction Technique FBP 30% ASIR 50% ASIR 70% ASIR 200 mAs 150 mAs 100 mAs 50 mAs 18.3 6 0.8 (17.5–19.1) 13.4 6 0.6 (12.8–14.0) 10.9 6 0.5 (10.4–11.4) 8.5 6 0.4 (7.9–8.9) 24.0 6 0.8 (23.2–24.8) 18.6 6 0.7 (17.9–19.4) 15.3 6 0.6 (14.7–15.9) 11.5 6 0.5 (11.0–12.0) 28.6 6 0.9 (27.7– 29.5) 21.5 6 0.7 (20.8– 22.3) 17.6 6 0.6 (17.0–18.2) 13.8 6 0.5 (13.3–14.3) 39.6 6 1.3 (38.3–41.1) 29.8 6 1.3 (28.5– 31.2) 23.1 6 0.9 (22.2–24.0) 18.9 6 0.7 (18.2–19.6) Note.—Data are mean objective image noise (in Hounsfield units) 6 standard deviation, with ranges in parentheses. Objective image noise decreased in the aorta with an increase in ASIR level of image reconstruction at all four tube current–time products. Table 6 Objective Image Noise and Modal Subjective Image Noise Scores for 22 Abdominal CT Studies according to Patient Weight Group and Image Reconstruction Technique Patient Weight Group, Tube Current–Time Product, and Type of Noise Score ⱕ90 kg 200 mAs Objective noise Subjective noise 150 mAs Objective noise Subjective noise 100 mAs Objective noise Subjective noise 50 mAs Objective noise Subjective noise .90 kg 200 mAs Objective noise Subjective noise 150 mAs Objective noise Subjective noise 100 mAs Objective noise Subjective noise 50 mAs Objective noise Subjective noise FBP 30% ASIR 50% ASIR 70% ASIR 13.3 6 2.6 2 9.9 6 2.0 2 8.4 6 1.5 1 6.4 6 1.5 1 17.6 6 4.1 4 14.2 6 3.0 3 11.6 6 2.3 2 8.8 6 2.0 2 32.1 6 4.8 3 17.3 6 3.8 3 14.1 6 2.6 2 10.8 6 2.4 1 30.7 6 6.1 5 24.7 6 5.1 4 19.7 6 4.8 3 15.3 6 3.8 3 13.9 6 4.0 2 10.8 6 3.6 2 10.7 6 2.8 1 8.5 6 2.7 1 18 6 4.7 3 16.2 6 2.0 3 13.1 6 2.1 2 10.6 6 2.2 1 22.9 6 5.6 3 19.9 6 3.0 2 16.3 6 3.2 2 13.1 6 2.9 1 31.2 6 7.8 5 27.7 6 3.3 4 22.3 6 3.9 3 17.3 6 3.9 3 Note.—Data are mean objective image noise (in Hounsfield units) 6 standard deviation or modal subjective image noise scores. Objective and subjective image noise was found to be lower in ASIR images than in FBP images for all four tube current–time products. (18); automatic exposure control (19); reducing the peak kilovoltage (20); using a higher pitch (21); and shielding radiosensitive organs such as the breast, thyroid, and lenses of the eye (22–24). The ASIR technique assessed Radiology: Volume 257: Number 2—November 2010 n in our study represents a new development for potential dose reduction. Our study shows that reduction of radiation dose down to 8.4 mGy is possible when abdominal CT images are reconstructed with 30% ASIR blending radiology.rsna.org and reduction of radiation dose down to 4.2 mGy is possible for patients weighing 90 kg or less with 50% and 70% ASIR blending. Regardless of the dose levels (4.2–16.8 mGy) and ASIR percentage levels, subjective image noise was consistently better with ASIR than with FBP. This trend in subjective image noise was supported by a corresponding decrease in quantitative noise in ASIR images as opposed to FBP images. Our results are consistent with the improved image noise described with the use of iterative reconstruction techniques to reconstruct phantom image data sets (6–10). These phantom studies have shown substantial improvement in image quality with iterative reconstruction compared with FBP for CT image reconstruction at low doses. Our results are also in agreement with the dose reduction reported in a recent phantom and patient study in which ASIR was used (25). In that smaller study, Hara et al reported that ASIR can provide diagnostic quality images at 32%–65% lower CTDIvol values than FBP techniques. Although both image noise and diagnostic confidence were acceptable on 50-mAs images reconstructed with 70% ASIR, a major change in image appearance due to substantial blotchy pixilation was noted at this level of blending of the ASIR and FBP techniques. Neither of the radiologists missed any lesions, despite the presence of these changes in image texture or appearance. Thus, although this blotchy, pixilated appearance did not interfere with lesion conspicuity and diagnostic confidence, it did generate a steplike artifact at tissue interfaces (such as the margins 381 GASTROINTESTINAL IMAGING: Reconstruction Techniques at Abdominal CT Singh et al Figure 5 Figure 5: Line graph summarizes MTF for 0.625-mm images CT reconstructed with FBP and ASIR at 30%, 50%, and 70% blending. Compared with the FBP images, ASIR images had higher spatial resolution at both 10% and 50% MTF, with the biggest improvement noted at 70% ASIR. lp/cm = Line pairs per centimeter, NO ASIR = FBP, STD = standard kernel. of the liver, spleen, and blood vessels). The exact reason for this appearance exclusively on ASIR images and not on FBP images is not known, but inherent differences in image reconstruction and a lack or paucity of image noise on ASIR images may have contributed to this appearance. Although quantitative image noise did increase with both FBP and ASIR with a decrease in dose and an increase in patient size, quantitative image noise with ASIR remained lower than with FBP for all dose levels in all patients, regardless of their weight or transverse diameter. This may be the reason that, regardless of patient size (weight = 62.8–97.4 kg) and transverse diameter (30.4–41.8 cm), ASIR images were found to be better in terms of image quality than FBP images in patients greater than 90 kg. 382 Our study had limitations. Foremost, our study had a small sample size given the difficulty in patient recruitment in the face of rising concern regarding radiation associated with CT scanning in the patient community. It is, however, possible that a larger study may provide paradoxical results in terms of other unknown artifacts or detection of lesions on ASIR images. Second, we did not investigate the effect of ASIR reconstruction in patients of different sizes. Presently, the ASIR reconstruction technique is available from one vendor and works only with that vendor’s images; therefore, we did not assess the effect of ASIR on images acquired with CT scanners manufactured by different vendors. Although it is very difficult to blind the experienced radiologist between ASIR and FBP images owing to differences in image texture, we randomized the image sets acquired at varying dose levels (4.2–16.8 mGy) and reconstruction techniques. Another possible limitation of our study was the fact that CT images were acquired in the equilibrium phase and not in an unenhanced or dynamic phase. While acquisition in an unenhanced phase would have limited our ability to detect subtle lesions and to focus the scanning acquisition to the location of the lesion, focused dynamic acquisition at four different dose levels was deemed impractical. Another consideration with our study was the fact that we used 16 image data sets in each patient at four radiation dose levels and reconstruction techniques, and this repetitive reviewing of images may have biased the radiologists in the lesion conspicuity assessment component of our study. However, to minimize this bias in image assessment, each radiologist was asked separately to first assess the lesion conspicuity on the image series with the greatest image noise or on those image data sets acquired at the lowest dose levels and then to assess the image data sets acquired at the other dose levels. Unfortunately, at present, this technique is available from one CT manufacturer only. A major implication of our study is the fact that reduction of the radiation dose can be achieved by lowering the x-ray tube current time– product when using the ASIR image reconstruction technique. Another implication of our study is the need to optimally set the correct ASIR-FBP blend. For a CTDIvol of 8.4 mGy (x-ray tube current time–product of 100 mAs), 30% and 50% ASIR blending is appropriate; further dose reduction will require a higher degree of ASIR blending, with the associated image texture alterations. Radiologists can reduce the radiation dose to 4.2 mGy by using 70% ASIR if they are willing to accept the possible induction of a blotchy, pixilated image appearance. In conclusion, reduction of CT radiation dose down to 8.4 mGy is feasible for abdominal CT images reconstructed with the ASIR technique without compromising image quality, lesion detection, and conspicuity. For patients weighing radiology.rsna.org n Radiology: Volume 257: Number 2—November 2010 GASTROINTESTINAL IMAGING: Reconstruction Techniques at Abdominal CT 90 kg or less, radiation dose reduction down to 4.2 mGy is possible with the ASIR technique. For routine abdominal CT examinations at 100 mAs in patients weighing greater than 90 kg, 30% and 50% ASIR-FBP blending provides acceptable image noise and diagnostic confidence, without a substantial change in image appearance. References 1. Shepp LA, Vardi Y. Maximum likelihood reconstruction for emission tomography. IEEE Trans Med Imaging 1982;1(2):113–122. 2. Lange K, Carson R. EM reconstruction algorithms for emission and transmission tomography. J Comput Assist Tomogr 1984;8(2): 306–316. 3. Kuhl DE, Edwards RQ. Image separation radioisotope scanning. Radiology 1963;80: 653–662. 4. Cormack AM. Representation of a function by its line integrals with some radiological applications. J Appl Phys 1963;34:2722–2727. nique (SART). IEEE Trans Image Process 2003;12(8):957–961. cal indication, and number of prior studies. Radiology 2009;252(1):200–208. 9. Wang J, Li T, Lu H, Liang Z. Penalized weighted least-squares approach to sinogram noise reduction and image reconstruction for low-dose X-ray computed tomography. IEEE Trans Med Imaging 2006;25(10): 1272–1283. 18. Cohnen M, Fischer H, Hamacher J, Lins E, Kötter R, Mödder U. CT of the head by use of reduced current and kilovoltage: relationship between image quality and dose reduction. AJNR Am J Neuroradiol 2000; 21(9):1654–1660. 10. Ziegler A, Köhler T, Proksa R. Noise and resolution in images reconstructed with FBP and OSC algorithms for CT. Med Phys 2007; 34(2):585–598. 19. Rizzo S, Kalra M, Schmidt B, et al. Comparison of angular and combined automatic tube current modulation techniques with constant tube current CT of the abdomen and pelvis. AJR Am J Roentgenol 2006; 186(3):673–679. 11. De Man B, Nuyts J, Dupont P, Marchal G, Suetens P. Reduction of metal streak artifacts in X-ray computed tomography using a transmission maximum a posteriori algorithm. IEEE Trans Nucl Sci 2000;47:977–981. 12. Wang G, Vannier MW, Cheng PC. Iterative X-ray cone-beam tomography for metal artifact reduction and local region reconstruction. Microsc Microanal 1999;5(1):58–65. 13. Robertson DD, Yuan J, Wang G, Vannier MW. Total hip prosthesis metal-artifact suppression using iterative deblurring reconstruction. J Comput Assist Tomogr 1997;21(2):293–298. 5. Kak AC. Image reconstruction from projections. In Ekstrom M, ed. Digital image processing techniques. New York, NY: Academic Press, 1984. 14. Silva AC, Lawder HJ, Hara A, Kujak J, Pavlicek W. Innovations in CT dose reduction strategy: application of the adaptive statistical iterative reconstruction algorithm [R2-2]. AJR Am J Roentgenol 2010;194(1): 191–199. 6. Snyder DL, Miller MI, Thomas LJ, Politte DG. Noise and edge artifacts in maximumlikelihood reconstructions for emission tomography. IEEE Trans Med Imaging 1987;6: 228–238. 15. Schindera ST, Nelson RC, Toth TL, et al. Effect of patient size on radiation dose for abdominal MDCT with automatic tube current modulation: phantom study. AJR Am J Roentgenol 2008;190(2):100-105. 7. Liow JS, Strother SC. Practical tradeoffs between noise, quantitation, and number of iterations for maximum likelihood-based reconstructions. IEEE Trans Med Imaging 1991;10(4):563–571. 16. EUR 16262. European guidelines on quality criteria for computed tomography. www.drs .dk/guidelines/ct/quality/download/eur16262 .w51. Accessed September 10, 2009. 8. Jiang M, Wang G. Convergence of the simultaneous algebraic reconstruction tech- Radiology: Volume 257: Number 2—November 2010 n Singh et al 17. Singh S, Kalra MK, Moore MA, et al. Dose reduction and compliance with pediatric CT protocols adapted to patient size, clini- radiology.rsna.org 20. Heyer CM, Mohr PS, Lemburg SP, Peters SA, Nicolas V. Image quality and radiation exposure at pulmonary CT angiography with 100or 120-kVp protocol: prospective randomized study. Radiology 2007;245(2):577–583. 21. Diel J, Perlmutter S, Venkataramanan N, Mueller R, Lane MJ, Katz DS. Unenhanced helical CT using increased pitch for suspected renal colic: an effective technique for radiation dose reduction? J Comput Assist Tomogr 2000;24(5):795–801. 22. Hidajat N, Schröder RJ, Vogl T, Schedel H, Felix R. The efficacy of lead shielding in patient dosage reduction in computed tomography [in German]. Rofo 1996;165(5):462–465. 23. Beaconsfield T, Nicholson R, Thornton A, Al-Kutoubi A. Would thyroid and breast shielding be beneficial in CT of the head? Eur Radiol 1998;8(4):664–667. 24. Hein E, Rogalla P, Klingebiel R, Hamm B. Low-dose CT of the paranasal sinuses with eye lens protection: effect on image quality and radiation dose. Eur Radiol 2002;12(7): 1693–1696. 25. 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):764–771. 383