Date of Award

5-10-2026

Date Published

June 2026

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Biomedical and Chemical Engineering

Advisor(s)

Zhen Ma

Keywords

Biology;Cardiac Modeling;Cardiotoxicity;Deep Learning;hiPSC-CM;Machine Learning

Abstract

Advances in stem cell technology have allowed researchers to create close proximities of in-vivo organs with advanced 3d tissue structures in vitro called organoids. These organoid models closely replicate the in-vivo physiology and structure of organs, allowing for the study of human biology, embryogenesis, disease and cytotoxicity in a more physiologically relevant context than traditional animal models. Specifically, advances in cardiac organoid research have resulted in functional spontaneously contracting cardiac models that can be used to investigate heart development and cardiotoxicity down to the level of a single cell. Interestingly, analytical tools have not been explored or developed at the same pace leading to an analytical bottleneck resulting in a lack of detailed insights from these extremely complex biological models. The goal of this dissertation is to investigate and apply advanced and novel analytical tools like machine learning, artificial intelligence and advanced analytical computational tools to the investigation of cardiac specific in-vitro models to further study cardiac physiology. Building upon models using either cardiac organoids or single cell hiPSC-CMs created with micropatterning techniques, we mine the structural and physiological functioning of the cells to elucidate the structure-function correlation of these cells. Utilizing novel tools from the field of computer science, we apply novel computational techniques beyond traditional linear analysis. By applying nonlinear analysis, machine learning, manifold learning and deep learning we gained deeper biological insights into the functioning of cardiac cells through their structure and cellular signaling. We outline the implementation of nonlinear analysis combined with machine learning as an extremely sensitive diagnostic tool for cardiotoxicity, allowing us to derive biological insights from chaos theory. Additionally, we found unsupervised machine learning tools can be used to mine higher dimensional data for patterns and insights that traditional analysis can miss in an unbiased manner. Combining supervised and unsupervised machine learning with network analysis can help investigate what drives organoid functioning and gives us the ability to link the functioning of organoids with their geometric structure in a novel analytical pipeline. Generative AI allows the prediction of cellular functioning from both cellular and subcellular structures of cardiomyocytes. Single-cell RNA sequencing allowed us to use unsupervised machine learning with transcriptomics to link the structure and functioning of cardiac organoids with their RNA signaling. This dissertation discusses and applies a wide range of novel computational tools to enhance our understanding of cardiac physiology for advanced investigations into cardiac physiology from the single cell to organoid level. Here, I have detailed the methodologies for investigating the optimization of organoids investigating the impact of structure on function, allowing us to predict and link physiological functioning from cellular and subcellular structure and cellular signaling.

Access

Open Access

Download

Available for download on Thursday, June 17, 2027

Share

COinS