Date of Award

5-2026

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Biomedical Engineering and Sciences

First Advisor

Linxia Gu

Second Advisor

Xianqi Li

Third Advisor

Pengfei Dong

Fourth Advisor

Peshala Gamage

Abstract

Percutaneous coronary intervention (PCI) is the dominant revascularization strategy for coronary artery disease (CAD), yet imaging-derived metrics exhibit limited sensitivity in predicting patient-specific procedural outcomes. Over the past few decades, significant advances have been made in understanding the mechanics of soft biological tissues, with computational biomechanics such as the finite element method (FEM) emerging as a powerful predictive tool for PCI. Nonetheless, several technical challenges—including the accurate 3D reconstruction from patients’ imaging, integration of constitutive failure models, and rapid computation of high-dimensional stress fields—continue to limit the clinical translation of FEM. In this dissertation, we addressed these challenges through the incorporation of detailed plaque morphology into patient-specific models, the development of a phase-field framework to represent fracture in calcified tissues, and the generation of a large-scale dataset of FEM-derived simulations, with the ultimate goal of training a biomechanics-informed artificial intelligence (AI) for PCI outcome predictions. Analyses of the FEM dataset revealed that lipid arc (r=0.769) along with area stenosis (r=0.550) and lumen curvature (r=0.642), were strong indicators for plaque rupture, with fibrous cap thickness showing a comparatively weak influence under interventional loading. The introduction of the phasefield model demonstrated that despite the reduced compliance of concentric calcified plaques, upon calcification fracture, they achieved a three-fold increase in lumen gain compared with the more eccentric cases (< 180◦ arc angle). All captured trends were fitted to a denoising diffusion model (DM) conditioned on labeled intravascular optical coherence (IVOCT) images, mapping the initial plaque morphology to stresses and deformed states of the post-PCI artery. Our efforts resulted in a structural similarity index > 0.894, learned perceptual image patch similarity < 0.048, and lumen gain deviation of 6.86% from FEM, indicating high-quality prediction of the stress field and deformation. The ending computational tool offers a rapid and reliable alternative to conventional FEM, supporting its integration into clinical workflows.

Download

Available for download on Monday, November 09, 2026

Share

COinS