Imran Shah
BME PhD Defense Presentation

Date: 2025-12-05
Time: 1:00 - 3:00 P.M.
Location / Meeting Link: UAW Room 3115, McIntire Conference Room (https://gatech.zoom.us/j/8942518357)

Committee Members:
Lakshmi Prasad Dasi, PhD (Advisor); Alessandro Veneziani, PhD (Co-Advisor); John Oshinski, PhD; Rudolph L. Gleason, PhD; Habib Samady, MD; Vinod Thourani, MD


Title: Real-Time Predictive Computational Modeling of Cardiovascular Interventions via Reduced Order Models

Abstract:
Predictive computational modeling has become a critical tool for pre-operative planning of complex cardiovascular procedures, allowing clinicians to predict potential post-operative risks and explore alternative strategies before entering the operating room. These tools have become especially popular in the context of transcatheter aortic valve replacement (TAVR), where device deployment simulations involving structural analysis of the prosthetic heart valve are performed, and Fontan surgical planning, where fluid dynamic simulations are performed to contextualize patient-specific hemodynamics and identify an optimized Total Cavopulmonary Connection (TCPC) configuration. These models are generally built upon traditional numerical techniques such as the Finite Element method to solve the underlying partial differential equations that govern the mechanical problem of interest (fluid or solid). However, due to their multiphyiscal nature, the computational costs required for these models become extremely expensive, rendering them impractical for routine clinical decision making. In this dissertation, we establish and validate a series of novel reduced order modeling (ROM) driven paradigms to bridge this gap by delivering real-time, patient-specific predictions with modest computational resources. In Specific Aim 1, projection-based ROMs are introduced for benchmark biomechanical problems in the TAVR and Fontan spaces, enabling real-time deformation simulations of TAVR prostheses and accelerated hemodynamic analyses of idealized 2D and 3D TCPC geometries. In Specific Aim 2, the idealized TAVR ROM framework is extended to simulate the entirety of the TAVR deployment process in patient-specific anatomies via a novel surrogate contact model, enabling real-time deployment simulations and rapid risk assessment of coronary obstruction post-TAVR. Finally, in Specific Aim 3 a novel ROM-driven Fontan surgical planning paradigm is introduced for patient-specific TCPC morphologies, combining geometrical parameterization techniques with the ROM for real-time hemodynamic and optimization studies of the TCPC. Collectively, these contributions demonstrate that reduced-order, physics-based computational models can provide real-time, clinically meaningful predictions in both TAVR and Fontan surgical planning using only local computational facilities often seen in the clinic (i.e., a simple laptop).