Date of Award
Fall 2023
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Biomedical Engineering (ENAS)
First Advisor
Duncan, James
Abstract
Accurate heart geometry modeling from medical images is important for both cardiovascular research and clinical practice. Prominent use cases include anatomical measurements, physics-based simulations, and 3D-printed phantoms. Various computational approaches have been proposed to automate the geometry modeling tasks. Most existing methods have utilized a ``bottom-up" approach, starting with a simpler geometry representation and progressively increasing its complexity to obtain the final desired output. Such methods have often consisted of mixtures of techniques that tackle each sub-problem separately, which can offer flexibility in the modeling process but can also lead to inefficient workflows and high dependency on human experts. In contrast, our main contributions utilize a ``top-down" approach, which leverages data-driven learning to directly predict the most complex geometry representation. From there, other geometrical information from simpler representations can be extracted robustly and automatically. Our main focus throughout the dissertation was on reliable automated algorithms for patient-specific volumetric mesh generation for the complex stenotic aortic valve geometry. Our proposed methods span a wide range of medical imaging-based heart tissue modeling approaches, such as segmentation, template deformation, and mesh optimization. Then, we proposed a robust calcification meshing algorithm that helps enhance the modeling capabilities of our final geometry. With the combined output, we performed various downstream analyses and demonstrated our techniques' applicability to patient-specific solid and fluid modeling of the heart.
Recommended Citation
Pak, Daniel Hyungseok, "Data-driven Heart Geometry Modeling for Patient-specific Biomechanics" (2023). Yale Graduate School of Arts and Sciences Dissertations. 1182.
https://elischolar.library.yale.edu/gsas_dissertations/1182
