Date of Award
5-14-2024
Degree Name
Doctor of Philosophy (PhD)
Degree Type
Dissertation
Abstract
Atherosclerosis is a leading cause of cardiovascular mortality worldwide, with specific carotid plaque components and inflammation within carotid arteries playing a critical role in plaque progression and vulnerability. To assess carotid plaque and help diagnose atherosclerosis, this dissertation illustrates deep learning-based automated methods to characterize plaque components on magnetic resonance (MR) images and to analyze inflammation biomarkers on positron emission tomography (PET) images. Moreover, this dissertation also investigates the correlation between plaque components on MR images and inflammation detected via PET imaging. The dissertation begins by illustrating a novel two-staged deep learning approach to segment and classify carotid artery atherosclerotic plaque components on multi-weighted MR images. Leveraging convolutional neural networks (CNN) and Bayesian neural networks (BNN), the proposed method integrates ground truth data from high-resolution ex vivo MRIs and histopathology, yielding accurate delineation of plaque components, specifically lipid-rich necrotic core (LRNC) with hemorrhage and calcification. Evaluation studies demonstrate superior performance compared to manual segmentation and existing deep learning methods, thereby enhancing patient risk assessment for ischemic stroke. Subsequently, the dissertation addresses inflammation detection within carotid plaques using PET imaging with 64Cu CANF-comb tracer. Overcoming challenges posed by partial volume effects (PVE), especially spillover effects, the proposed end-to-end deep learning approach, enhanced with a customized multiscale Residual backbone (MSR) and a novel simulation pipeline, achieves precise inflammation segmentation and uptake quantification. Evaluation results showcase the accuracy of method, offering a promising tool for early detection of vulnerable plaques and prevention of cerebrovascular events. Furthermore, this thesis investigates the correlation between carotid plaque components identified on MR images and inflammation quantified on PET images. Through comprehensive analysis and statistical evaluation, we reveal the relationship between specific carotid plaque components and inflammation biomarkers, providing insights into the interplay between carotid plaque components and inflammation within carotid arteries. By integrating advanced medical imaging techniques and deep learning methods, this dissertation contributes to a deeper understanding of carotid plaque components and inflammation analysis, with significances for future personalized risk assessment of stroke events and therapeutic strategies in ischemic stroke prevention.
Language
English (en)
Chair
Jie Zheng
Committee Members
Pamela Woodard