Feature-based characterization and image quality optimization of motion-contaminated coronary structures in cardiac CT images.

摘要

In cardiac multidetector computed tomography (CT), the continuous motion of the beating heart can cause artifacts in the reconstructed images. These motion artifacts can adversely affect the accurate delineation of critical coronary structures, which may show indications of coronary artery disease. Computerized methods potentially can aid physicians in the interpretation of these motion-contaminated images, by allowing for a quantitative analysis of these images in an automated, efficient, and consistent manner. In this dissertation, a unified framework that explicitly accounts for patient-specific, reconstruction-specific, and lesion-specific information in the characterization and optimization of coronary structures in cardiac CT images is presented.; The initial study was focused on suppressing motion artifacts affecting calcified plaques by using the novel backprojection filtration algorithm, which can reconstruct region-of-interest images with less data than those required by conventional algorithms. The focus then shifted toward the characterization of motion artifacts affecting calcified plaques in noncontrast-enhanced cardiac CT images. An automated method was developed for extracting image-based features of motion-contaminated plaques at phases throughout the cardiac cycle.; The image-based features of calcified plaques formed the basis for three additional studies regarding the interpretation of cardiac CT images. In two studies, extracted features as well as patient-specific and reconstruction-specific parameters were inputted into regression models for predicting assessability indices reflective of image quality. These studies, which were conducted on both software and hardware phantoms, demonstrated that the variability of assessability indices predicted by these models lie within the variability of indices assigned by expert physicians. In the third study, a statistical analysis of the association between assessability indices and errors in coronary calcium score was performed using the previously acquired phantom data.; A unified framework for optimizing cardiac CT image quality was then developed. In this framework, image quality metrics were used to guide the selection of reconstruction parameters that improved the image quality of coronary structures. This study fully demonstrates the benefit of merging image reconstruction and image analysis techniques in cardiac CT research.