Date of Award

5-2026

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Mechanical Engineering

Committee Chair/Advisor

Ethan Kung

Committee Member

Richard Figliola

Committee Member

Georges Fadel

Committee Member

Gregory Mocko

Committee Member

Yuhao Xu

Abstract

Cardiovascular diseases accounted for approximately 19.5 million deaths in 2020, making them the leading cause of non-communicable mortality worldwide. This global burden underscores the need for improved tools to study, diagnose, and treat cardiovascular conditions. Mathematical modeling provides a means to investigate cardiovascular physiology and for preliminary assessment of novel interventions under reproducible conditions.

Recent developments in computational modeling, such as Partitioned Multiscale Methods (PMMs), have enabled the integration of two extensively utilized approaches: Lumped Parameter Network (LPN) models and three-dimensional (3D) finite element simulations. LPN models are widely employed due to their capacity to reproduce physiologically realistic pressure and flow dynamics at minimal computational cost. In contrast, 3D finite element models provide detailed spatial and temporal representations of blood flow, though at a significantly greater computational cost. PMMs establish a bidirectional exchange of information between the LPN and 3D domains at each timestep, and facilitate dynamic coupling between global (LPN) and local (3D finite element) hemodynamic phenomena. The combined output of a PMM is a synthesis of both domains allowing the simultaneous evaluation of both global information such as cardiac output or vena caval pressures, and local information such as energy loss.

The first project of this dissertation employs a PMM framework to inform surgical planning for patients with univentricular physiology. These patients undergo a series of staged palliative interventions culminating in the Fontan procedure. Previous studies suggest that the energetic efficiency of the Fontan surgical geometry is correlated with exercise capacity. This has led to efforts at optimizing the Fontan surgical geometry to minimize localized junction energy loss, as a means to enhancing long term health and exercise capacity, thereby improving quality of life. Using a validated PMM framework and an existing automated tuning protocol for a closed-loop LPN representing Fontan physiology, six patients were evaluated, each with three distinct junction geometries and six exercise stages, resulting in 108 simulations. These analyses enabled simultaneous evaluation of global metrics, such as cardiac output, and local parameters, such as junction energy loss. The performance of the three junction geometries were then assessed in terms of both localized and global metrics, and our findings were translated into recommendations for surgical planning.

However, conducting a clinical application study such as the first project for the wider population with biventricular physiology is currently unfeasible. A major limitation of PMM remains the labor-intensive process of tuning LPN parameters, especially for complex closed-loop models. The existing automated LPN tuning protocol is only applicable to the Fontan physiology only, which constitutes a minor fraction (66 Fontan patients per million) of the general population. Therefore there is a clear need for an automated tuning protocol for a biventricular LPN, which the second project of this dissertation addresses. We tested the proposed protocol against 500 physiologically realistic hemodynamic target sets. The results demonstrated that the protocol accurately matches multiple hemodynamic targets such as cardiac output and aortic pressures, while maintaining realistic atrial and pulmonary pressures. The proposed protocol is a reproducible, adaptable tool that significantly reduces manual tuning requirements, and supports broader translational research.

Author ORCID Identifier

0000-0003-2858-4039

Download

Available for download on Monday, May 31, 2027

Share

COinS