Assessment of a model-based, iterative reconstruction algorithm (MBIR) regarding image quality and dose reduction in liver computed tomography

摘要

Objectives: The purpose of this study was to assess the image quality of half-dose (HD) liver computed tomography (CT) using a model-based iterative reconstruction algorithm (MBIR) compared with reference dose (RD) using filtered back projection (FBP) and the HD CT images using FBP and adaptive statistical iterative reconstruction (ASIR). Materials and Methods: A total of 103 patients suspected of having liver metastases underwent liver CT including HD portal venous phase imaging. Among these patients, 73 had undergone RD liver CT reconstructed using FBP, and the HD portal phase CT scans were separately reconstructed using FBP and MBIR. For the other 30 patients, the HD CT images were reconstructed using FBP, ASIR, and MBIR. The CT attenuation coefficients and the mean image noise of various sites, including the liver, the aorta, the main portal vein (MPV), and the subcutaneous fat, weremeasured, and the contrast-to-noise ratios (CNRs) of the metastatic lesion to the liver and theMPV to the liver were calculated. Two radiologists reviewed each image set with regard to image noise, image quality, lesion conspicuity, and diagnostic acceptability. Results: Compared with RD CT, there was a 46.1% decrease in CT dose index volume with HD CT. Image noise was significantly lower in the HD images reconstructed with MBIR than in both the HD FBP images and the RD FBP images(P < 0.001). Compared with the RD FBP and HD FBP images, the CNRs of the metastatic lesion to the liver and the MPV to the liver were higher in the HD MBIR images(P < 0.001). Despite the presence of the unique whirling artifacts of the MBIR images, the HD MBIR images were of good to excellent quality and were not inferior to RD FBP images regarding the lesion conspicuity, the image quality, and the diagnostic acceptability (P > 0.05). Half-dose MBIR also showed less image noise, higher CNRs, and superior image quality compared with HD ASIR(P < 0.001). Conclusions: The HDMBIR images showed less noise, higher CNR, and better image quality than the HDASIR andHDFBP images did; they also provided less image noise, higher CNR, and similar image quality compared with those of RD FBP images.