Improved Image Quality in Dual- Energy Abdominal CT: Comparison of Iterative Reconstruction in Image Space and Filtered Back Projection Reconstruction

Transcription

1 Medical Physics and Informatics Original Research Wang et al. Image Reconstruction Algorithms for Dual-Energy Abdominal CT Medical Physics and Informatics Original Research Rui Wang 1 Wei Yu 1 Runze Wu 2 Hua Yang 3 Dongxu Lu 1 Jiayi Liu 1 Zhaoqi Zhang 1 Chuanchen Zhang 1 Wang R, Yu W, Wu R, et al. Keywords: abdomen, dual-energy CT, image quality, iterative reconstruction DOI: /AJR Received April 27, 2011; accepted after revision December 21, R. Wu is a consultant at Siemens Healthcare. 1 Department of Radiology, Beijing Anzhen Hospital, Capital Medical University, 2 Anzhen Rd, Beijing, China. Address correspondence to Z. Zhang (zhaoqi5000@vip.sohu.com). 2 Siemens Healthcare China, Beijing, China. 3 Affiliated Hospital of Hebei United University, Tangshan, Hebei, China. AJR 2012; 199: X/12/ American Roentgen Ray Society Improved Image Quality in Dual- Energy Abdominal CT: Comparison of Iterative Reconstruction in Image Space and Filtered Back Projection Reconstruction OBJECTIVE. The purpose of this study is to investigate whether an iterative reconstruction in image space (IRIS) algorithm improves the image quality of dual-energy CT abdominal examinations performed during the hepatic arterial phase. MATERIALS AND METHODS. Seventy patients with suspected liver masses underwent contrast-enhanced multiphase abdominal examination and were enrolled in the study. A dual-energy CT protocol was performed in the hepatic arterial phase (parameters: tube A, 140 kv and 90 ma; tube B, 80 kv and 382 ma; automatic tube current modulation on; and collimation, mm). The reconstructions were performed with filtered back projection (FBP) and IRIS algorithms at a slice thickness of 3 mm and kernels of B30 and I30. The image noise was measured on the liver, aorta, and subcutaneous fat on the FBP and IRIS fusion images (m = 0.3) at the same position. The image noise and diagnostic acceptability of all images were scored by two radiologists. RESULTS. The image noise using the IRIS algorithm was lower than that using the standard FBP algorithm on the liver, aorta, and subcutaneous fat, respectively. The signal-tonoise ratio and contrast-to-noise ratio of images reconstructed with the IRIS algorithm also were significantly higher than for those reconstructed with the FBP algorithm. The diagnostic acceptability score using the IRIS algorithm was higher than that using the FBP algorithm at the same dose level (1.20 ± 0.40 vs 1.37 ± 0.57; p < 0.05). CONCLUSION. Compared with standard FBP reconstruction, an IRIS algorithm enables significant reduction of image noise and improvement of image quality and has the potential to decrease radiation exposure during contrast-enhanced dual-energy CT abdominal examination. D ual-energy CT provides the capability to differentiate between different materials at imaging [1]. This capability enables multiple dual-energy CT applications, including the triage of patients with nephrolithiasis on the basis of stone composition [2 4], evaluation of tissue perfusion blood volume [5, 6], the possibility of eliminating a separate unenhanced CT acquisition [7], and improved detection of pathologic hyperenhancement [8, 9]. Dual-energy CT acquisition includes images obtained at two different tube voltages: a high and a low peak kilovoltage. A fusion image of images obtained at two different peak kilovoltages is routinely generated to provide a 120-kVp-equivalent image for clinical use. However, the use of the low-peak-kilovoltage image frequently causes an increase in image noise and beam-hardening artifacts. This may result in a loss of image quality and problems in differential diagnosis. Recently, iterative reconstruction has been reintroduced to reduce radiation dose and improve image quality [10 12]. It has been shown that an iterative reconstruction in image space (IRIS) algorithm enables a significant reduction in image noise without a loss of diagnostic information for chest CT examinations [13]. However, to our knowledge, the IRIS algorithm has not been studied in dual-energy abdominal CT examinations. Therefore, the purpose of the current study was to determine the performance of the IRIS algorithm in improving the quality of fusion images of dual-energy CT abdominal examinations performed during the hepatic arterial phase. Materials and Methods Patient Population From June 2010 to July 2011, 70 patients (28 women and 42 men) with known or suspected liver masses referred for multiphase contrast-enhanced 402 AJR:199, August 2012

2 Image Reconstruction Algorithms for Dual-Energy Abdominal CT abdominal CT examination were retrospectively enrolled in this study. The mean (± SD) age of patients was 60 ± 10 years (range, years), the mean height was ± 7.47 cm (range, cm), the mean weight was ± kg (range, kg), and the mean body mass index was ± 2.90 kg/m 2 (range, kg/ m 2 ). Exclusion criteria were contraindications to iodinated contrast media, overt heart failure (New York Heart Association class III or IV), and renal insufficiency (serum creatine level > 1.4 mg/ dl). The study was approved by the institutional review board, and informed consent was obtained from all participating patients. Abdominal Contrast-Enhanced CT Scan Technique Multiphase abdominal CT was performed on a dual-source CT scanner (SOMATOM Definition, Siemens Healthcare). A dual-energy CT protocol was performed in the arterial phase. The parameters included helical scanning mode (tube A, 140 kv and 90 ma; tube B, 80 kv and 382 ma), automatic tube current modulation (CARE Dose4D, Siemens Healthcare), collimation of mm, pitch of 0.55, and gantry rotation time of 0.5 second. For triggering of the arterial phase acquisition, a bolus-tracking technique was used by placing a region of interest (ROI) at the supraceliac artery, and image acquisition was automatically triggered 12 seconds after the signal attenuation reached the predefined threshold of 100 HU [14]. The venous and equilibrium phases were at fixed time delays of 70 and 180 seconds, respectively. The venous and equilibrium phases were acquired during single-breath-hold helical scanning of the abdomen by using our routine protocol at a tube voltage of 120 kv, a tube current of 190 ma, and automatic tube current modulation. These sets were not included in the quantitative analyses because they were single-energy acquisition data. The volume CT dose index and doselength product were given by the CT scanner, and the effective dose was calculated by multiplying the dose-length product by the conversion coefficient (0.015 msv mgy 1 cm 1 ), leading to an effective dose for dual-energy CT of 3.39 ± 0.84 msv during the hepatic arterial phase. The contrast agent was injected with a dualhead power injector (Stellant D, Medrad) using an 18-gauge IV needle placed in the right antecubital vein. A routine injection protocol was used. According to patient weight, ml/kg of contrast agent (370 mg I/mL iopromide; Ultravist, Bayer Healthcare) was injected, followed by 30 ml of saline (0.9% sodium chloride) as a bolus chaser. The total volume of contrast agent was ml. The injection rate was ml/s. Dual-Energy CT Image Reconstruction The 140- and 80-kVp images and a fusion image (140- and 80-kVp images mixed with a weight factor of 0.3) were reconstructed by using a standard FBP algorithm with 3-mm slice thickness, 1.5-mm increment, and a medium-smooth kernel (D30f). The raw data were further transferred to an offline workstation for IRIS reconstruction. The IRIS images were reconstructed with the same parameters as the FBP algorithm and a Q30f kernel, which matches the resolution of a D30f kernel. In this study, images using five iterations were reconstructed, because image quality was better after five iterations, as described by Pontana et al. [13]. Image Quality Analysis The FBP and IRIS fusion images were loaded into a commercial workstation to evaluate objective and subjective image quality. The two fusion image sets obtained using FBP and IRIS for each patient were displayed side by side with a preset window (window width, 400 HU; window level, 60 HU). According to the European guidelines on quality criteria for CT for abdominal CT examinations [15], two radiologists independently assessed the subjective image quality of all 70 dual-energy CT datasets (70 image sets with FBP and 70 image sets with IRIS) on the workstation over a period of 4 weeks. Visually sharp abdominal structures, image noise, and diagnostic acceptability were assessed by two radiologists. According to the guidelines, each radiologist assessed critical reproduction of visually sharp abdominal structures, including the liver parenchyma and intrahepatic portal veins, liver veins, structures of liver hilum, common hepatic and common bile ducts, gallbladder wall, and aorta. Image noise was assessed on a 3-point scale, where grade 1 meant too little or less than usual noise, grade 2 meant optimum noise, and grade 3 was given for CT examinations with too much noise affecting the image interpretation. Two radiologists assessed independently the overall image diagnostic acceptability (1, excellent; 2, good; 3, acceptable; and 4, unacceptable), according to a procedure described elsewhere [11]. Interobserver disagreements were resolved by consensus in a joint session. Objective Analysis The image noise (SD of CT number) of the liver, aorta, and subcutaneous fat was measured by manually placing circular ROIs at the same image level. The SD of liver was recorded as the mean of measurement of three ROIs, not including any vessel and artifacts in the ROI area. When the SD of the aorta was measured, calcification and plaque on the aortic wall were avoided. The SD of subcutaneous fat was recorded as the mean measurement of three ROIs placed in the anterior, posterior, and lateral abdominal wall. For all measurements, the size, shape, and position of the ROIs were kept constant among the two protocols (Fig. 1). For each protocol, the signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were calculated with the following formulas: SNR = ROIm / SDm and CNR = (ROIm ROI fat ) / SD fat, where ROIm is the mean attenuation of liver, aorta, and fat tissue; ROI fat is the mean attenuation of subcutaneous fat; SDm is the mean of the image noise at the liver and aorta; and SD fat is the mean of the subcutaneous fat noise. Statistical Analysis Statistical analysis was performed using SPSS software (version 16.0, SPSS). Continuous variables were described as mean ± SD. The image noise (SD), SNR, and CNR of the FBP and IRIS groups were compared using paired Student t tests. Nonparametric tests (McNemar chi-square test) were used to compare subjective scores of image noise and diagnostic acceptability. A kappa test was used to check the agreement between the results of the two radiologists for subjective quality. The kappa statistics were weighted as follows: 0 0.2, low; , moderate; good, , substantial, and greater than 0.81, perfect agreement, and p values less than 0.05 were considered to be statistically significant. Results Subjective Image Quality There was moderate interobserver agreement between the two radiologists for subjective image quality assessed in all 70 dualenergy abdominal CT image sets (κ = 0.726; p < 0.05). Visually sharp abdominal structures, image noise, and diagnostic acceptability with FBP and IRIS are summarized in Table 1 (see also Figs. 1 and 2). There was a significant reduction in the level of subjective image noise overall for IRIS images compared with FBP images (p < 0.05); furthermore, the diagnostic acceptability score for IRIS images was higher than that for FBP images (1.20 ± 0.40 vs 1.37 ± 0.57; p < 0.05). Objective Image Quality Our results showed that, when the radiation dose was kept constant, the IRIS algorithm yielded significantly decreased image noise for the aorta, liver, and subcutaneous fat, respectively, and increased SNR and CNR for the aorta and liver and increased SNR for subcutaneous fat, respectively, compared with the standard FBP algorithm (all p < 0.001; Table AJR:199, August

3 Wang et al. 2 and Figs. 1 and 2). However, there was no statistically significant difference in CT values between IRIS and FBP images (all p > 0.05; Table 2). Discussion To our knowledge, this is the first study to evaluate the performance of an IRIS algorithm in abdominal dual-energy CT. Compared with FBP and IRIS imaging from TABLE 1: Comparison of Subjective Image Quality Between Filtered Back Projection (FBP) and Iterative Reconstruction in Image Space (IRIS) Algorithms Variable FBP IRIS Visual sharpness Positive Positive Image noise a Diagnostic acceptability b Note Data are no. of images assigned a particular grade for image noise and diagnostic acceptability. a Image noise was graded on a 3-point scale: grade 1, too little or less than usual noise; grade 2, optimum noise; and grade 3, too much noise affecting the image interpretation. b Overall image diagnostic acceptability was graded on a 4-point scale: 1, excellent; 2, good; 3, acceptable; and 4, unacceptable. Fig year-old woman. A and B, Axial contrast-enhanced dual-energy CT images obtained during hepatic arterial phase processed with iterative reconstruction in image space (A) and filtered back projection (B) algorithms show regions of interest manually drawn on aorta, liver, and subcutaneous fat on same level. Fig year-old woman with small hepatic cavernous hemangioma and two cysts. A and B, Contrast-enhanced dual-energy CT images obtained during hepatic arterial phase processed with iterative reconstruction in image space (IRIS) (A) and filtered back projection (FBP) (B) algorithms show significantly reduced image noise and improved image quality with IRIS compared with FBP. IRIS can enhance spatial resolution at edge of highcontrast tissues (arrowheads), maintaining boundary sharpness. Small vessel near cystic lesion (arrow) is shown clearly on IRIS compared with FBP. A A the same datasets, our results show that the IRIS algorithm significantly decreases image noise and improves image quality for abdominal dual-energy CT fusion images obtained during the hepatic arterial phase. A dual-energy CT scan generates three datasets by default: high peak kilovoltage, low peak kilovoltage, and a fusion image. The fusion image is a linear mix of the highand low-peak-kilovoltage images and is usually considered a 120-kVp-equivalent image. These images are routinely used for the purpose of diagnosis and the starting point for advanced dual-energy postprocessing. Therefore, maintaining sufficient image quality of the fusion image is desired. The IRIS algorithm introduces a loop of image correction in the image space, which iteratively reduces image noise while maintaining boundary sharpness [13]. It was found that IRIS can enhance spatial resolution at the edge of high-contrast tissues and reduce image noise on homogeneous areas at the same radiation dose [13]. In our study, 80- and 140-kVp image data, the fusion images, and virtual image data were reconstructed using the IRIS algorithm. Consistent with the results of previous studies [10 13, 16, 17], the image quality of these image sets was significantly improved by using IRIS instead of FBP. In the present study, IRIS reduced the image noise better than FBP did at the same doses level (p < 0.001). At the same time, IRIS yielded significantly higher SNR and CNR on the liver parenchyma and aorta than did FBP (p < 0.001). Clinically, it is vitally important that subjective assessments of image quality by radiologists are improved through various reconstruction algorithms. Our results show that image noise and diagnostic acceptability of images were significantly improved with the IRIS reconstruction algorithm over the FBP B B 404 AJR:199, August 2012

4 Image Reconstruction Algorithms for Dual-Energy Abdominal CT TABLE 2: Comparison of Objective Image Quality Between Filtered Back Projection (FBP) and Iterative Reconstruction in Image Space (IRIS) Algorithms 140-kV Image 80-kV Image Fusion Image a Virtual Unenhanced Image Variable FBP IRIS FBP IRIS FBP IRIS FBP IRIS p Liver CT value ± ± ± ± ± ± ± ± 5.11 > 0.05 Liver noise (HU) ± ± ± ± ± ± ± ± 1.07 < Liver SNR 3.29 ± ± ± ± ± ± ± ± 1.67 < Liver CNR ± ± ± ± ± ± ± ± 8.94 < Aorta CT value ± ± ± ± ± ± ± ± > 0.05 Aorta noise (HU) ± ± ± ± ± ± ± ± 1.76 < Aorta SNR ± ± ± ± ± ± ± ± 1.92 < Aorta CNR ± ± ± ± ± ± ± ± 8.59 < Fat noise (HU) ± ± ± ± ± ± v ± 2.59 < Fat SNR 6.17 ± ± ± ± ± ± ± ± 2.59 < Note Data are mean ± SD. CNR = contrast-to-noise ratio, HU = Hounsfield units, SNR = signal-to-noise ratio. a Fusion image is created by merging 140- and 80-kV datasets (m = 0.3) to yield a virtual 120-kV image. algorithm. A comparison of the two reconstruction algorithms showed better boundary sharpness of lesions on IRIS images, confirming that reduction of the objective image noise reinforces the diagnostic accuracy. Although the so-called plastic image impression in iterative reconstruction has been reported [18], the images reconstructed with IRIS are easily appreciated. However, the time needed for an iterative reconstruction was significantly longer than that needed for FBP reconstruction. The next-generation iterative algorithm with hardware acceleration will be desired for routine clinical usage. Our findings support that IRIS has important clinical applications. Compared with the standard FBP algorithm at the same level of radiation dose, the IRIS algorithm can reduce image noise and improve the image quality. Alternatively, the application of the IRIS technique could be another exercise for decreasing or optimizing radiation doses without losing the overall image quality [17] because the radiation dose of CT examinations and its associated risks are major concerns [19]. Thus, strategies aimed at lowering radiation dose have become particularly important. There were several limitations to this study. First, compared with standard FBP, IRIS significantly improved image noise and image quality, but these results reflect only a preliminary experience with a relatively small population. Second, we limited our assessment to the effect of IRIS I30 on fusion images of dual-energy CT. We did not assess whether alternative iteration would have resulted in substantially different results. In future work, we will evaluate a wide range of iterations on fusion image quality. Third, the assessment of the IRIS technique in this study, comparing it with standard FBP, was based on the same patients and the same levels of radiation dose. Future investigations of the effectiveness of new iterative reconstruction will be applied to reduced radiation doses, comparing the IRIS technique with standard FBP. In summary, this study shows that image quality is significantly improved in dual-energy CT abdominal examination using the IRIS algorithm compared with standard FBP. IRIS can be routinely applied on these studies, potentially reducing radiation exposure. Acknowledgments We thank the technical staff members of the CT department for invaluable contributions, particularly Hong Jiang, Weihua Zhang, Honghong Tie, Zhimin Li, Miao Guo, and Haixia Yang for technical assistance. References 1. Johnson TR, Krauss B, Sedlmair M, et al. Material differentiation by dual energy CT: initial experience. Eur Radiol 2007; 17: Takahashi N, Hartman RP, Vrtiska TJ, et al. Dualenergy CT iodine-subtraction virtual unenhanced technique to detect urinary stones in an iodinefilled collecting system: a phantom study. AJR 2008; 190: Graser A, Johnson TR, Bader M, et al. Dual energy CT characterization of urinary calculi: initial in vitro and clinical experience. Invest Radiol 2008; 43: Primak AN, Fletcher JG, Vrtiska TJ, et al. Noninvasive differentiation of uric acid versus nonuric acid kidney stones using dual-energy CT. Acad Radiol 2007; 14: Thieme SF, Becker CR, Hacker M, Nikolaou K, Reiser MF, Johnson TR. Dual energy CT for the assessment of lung perfusion: correlation to scintigraphy. Eur J Radiol 2008; 68: Ruzsics B, Lee H, Zwerner PL, Gebregziabher M, Costello P, Schoepf UJ. Dual-energy CT of the heart for diagnosing coronary artery stenosis and myocardial ischemia: initial experience. Eur Radiol 2008; 18: Stolzmann P, Frauenfelder T, Pfammatter T, et al. Endoleaks after endovascular abdominal aortic aneurysm repair: detection with dual-energy dual-source CT. Radiology 2008; 249: Schindera ST, Nelson RC, Mukundan S Jr, et al. Hypervascular liver tumors: low tube voltage, high tube current multi-detector row CT for enhanced detection phantom study. Radiology 2008; 246: Graser A, Johnson TR, Chandarana H, Macari M. Dual energy CT: preliminary observations and potential clinical applications in the abdomen. Eur Radiol 2009; 19: Marin D, Nelson RC, Schindera ST, et al. Lowtube-voltage, high-tube-current multidetector abdominal CT: improved image quality and decreased radiation dose with adaptive statistical iterative reconstruction algorithm initial clinical experience. Radiology 2010; 254: Prakash P, Kalra MK, Kambadakone AK, et al. Reducing abdominal CT radiation dose with adaptive statistical iterative reconstruction technique. Invest Radiol 2010; 45: Singh S, Kalra MK, Hsieh J, et al. Abdominal CT: comparison of adaptive statistical iterative and fil- AJR:199, August

5 Wang et al. tered back projection reconstruction techniques. Radiology 2010; 257: Pontana F, Pagniez J, Flohr T, et al. Chest computed tomography using iterative reconstruction vs filtered back projection. Part 1. Evaluation of image noise reduction in 32 patients. Eur Radiol 2011; 21: Sultana S, Awai K, Nakayama Y, et al. Hypervascular hepatocellular carcinomas: bolus tracking with a 40-detector CT scanner to time arterial phase imaging. Radiology 2007; 243: Menzel HG, Schibilla H, Teunen D, eds; European Commission s Radiation Protection Actions. Quality criteria for computed tomography. EUR Danish Society of Radiology Web site. www. drs.dk/guidelines/ct/quality/download/eur w51. Accessed December 29, 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: Pontana F, Duhamel A, Pagniez J, et al. Chest computed tomography using iterative reconstruction vs filtered back projection (Part 2): image quality of low-dose CT examinations in 80 patients. Eur Radiol 2011; 21: Leipsic J, Labounty TM, Heilbron B, et al. Adaptive statistical iterative reconstruction: assessment of image noise and image quality in coronary CT angiography. AJR 2010; 195: Brenner DJ, Hall EJ. Computed tomography: an increasing source of radiation exposure. N Engl J Med 2007; 357: FOR YOUR INFORMATION Malpractice Issues in Radiology, 3rd edition, by Leonard Berlin, is now available! For more information or to purchase a copy, see AJR:199, August 2012