Event Actions
Pixels-to-Profiles: Data-driven approaches to extract insights from diverse plaque phenotypes for atherosclerosis
Abstract:
Atherosclerosis remains the leading driver of heart attack and stroke, yet the plaque features that foreshadow clinical events, such as intraplaque hemorrhage (bleeding into the plaque), fibrous cap weakening, lipid-rich cores, calcification, and inflammation which are still assessed inconsistently across centers. This dissertation centers on cardiovascular biology and patient impact: I assemble a large, well-curated carotid endarterectomy cohort and show how systematic, imaging-based plaque phenotyping can make these high-risk features more objective, reproducible, and scalable. A particular emphasis is placed on intraplaque hemorrhage, a hallmark of unstable lesions linked to downstream ischemic events. In parallel, I connect plaque morphology to underlying biology by estimating where key gene programs are active across each slide, offering spatial context for processes like macrophage-driven inflammation, matrix remodeling, and smooth-muscle changes in the fibrous cap. Together, these advances deliver a standardized IPH phenotyping readout, uncover spatially organized gene programs linked to plaque instability, and produce concrete, biology-driven hypotheses for targeted validation in cardiovascular cohorts. The work is designed for practical use in cardiovascular research pipelines and has potential to support earlier risk stratification, mechanistic insight into plaque progression, and more precise therapy selection across vascular beds.
Committee:
- Miaomiao Zhang, Committee Chair, ECE, CS, SEAS/UVA
- Clint L. Miller, Advisor, Department of Genome Sciences, School of Medicine/UVA
- Yangfeng Ji, CS/SEAS/UVA
- Gustavo Rohde, BME,EE/SEAS/UVA
- Chongzhi Zang, Department of Genome Sciences, BME, and Biochemistry and Molecular Genetics/SEAS/UVA
- Sander W van der Laan, University Medical Center Utrecht and Utrecht University, Netherlands