Coronary atherosclerosis is by far the most frequent cause of ischemic heart disease. Intravascular ultrasound (IVUS) along with virtual histology (VH) is a useful tool for quantification of coronary plaque buildup and provides new insights into the diagnosis of coronary disease. Rupture of vulnerable plaque causing acute coronary syndromes, coronary remodeling maintaining lumen size and plaque phenotype revealing pathological severity are among the most important topics related to atherosclerosis. In this thesis, variations of IVUS-VH-derived thin-cap fibroatheroma (TCFA) definitions are proposed to evaluate the plaque rupture, which is further analyzed in a layered manner; statins effects on coronary remodeling are comprehensively assessed with the implementation of automated IVUS segmentation and registration of IVUS pullbacks based on baseline and 1-year followup datasets; plaque phenotypes are determined and analyzed morphologically and compositionally on segmental basis using the same serial datasets.
In addition, our research involves another important coronary disease — coronary allograft vasculopathy (CAV) which is a frequent complication of heart transplantation (HTx). Another intra-coronary imaging modality — intravascular optical coherence tomography (IVOCT) for quantifying CAV is involved. We present an optimal and automated 3-D graph search approach for the simultaneous IVOCT multi-layer segmentation by transforming the 3-D segmentation problem into finding a minimum-cost closed set in a weighted graph. Furthermore, a computer-aided just-enough-interaction refinement method is proposed to help achieve fully satisfactory 3-D segmentation of IVOCT images. We believe this is the first work that provides a fast, efficient and accurate solution for IVOCT multi-layer assessment in the context of CAV.
The major contributions of this thesis are: (1) Proving that IVUS-VH-derived TCFA prevalence may be overestimated, and elucidating the potential loss of plaque material during rupture, (2) providing a comprehensive understanding of remodeling in the context of both changing the remodeling direction and changing the remodeling extent, and demonstrating the statin therapy effects on remodeling across patients, based on automated segmentation of IVUS images and registration of serial data, (3) showing that the pathological intimal thickening is the most active plaque phenotype in terms of plaque composition changes and plaque vulnerability progression, and (4) developing and validating a method for multi-layer 3-D segmentation of IVOCT images within a novel interactive environment.