Data-driven approaches for complex systems: leveraging machine learning, materials science, and manufacturing for new biomedical technologies

Many research efforts to advance human health and well-being involve interdisciplinary problem spaces and complex, poorly-understood systems. This thesis integrates both computational and experimental approaches to advance our understanding and control of complex systems at the interface of machine learning, materials science, and manufacturing. Specifically, I demonstrate the data-driven description of supervised machine learning for biomedical engineering tasks, the data-driven design of optimized soft granular biomaterials, and the proof-of-concept development of a transcatheter additive manufacturing platform. In Part 1, I develop custom software for high-resolution, multifactorial machine learning (ML) experiments. I iteratively apply this workflow to a set of diverse ML problems from the biomedical engineering (BME) domain to generate massive meta-datasets covering each phase of the hierarchical ML optimization and evaluation process. Then, I describe the underlying patterns and heterogeneity in these rich datasets and delineate empirical guidelines for the rigorous and reliable adoption of machine learning for BME problems. In Part 2, I leverage the insights from Part 1 to develop a flexible and robust data-driven modeling pipeline for complex soft materials. The pipeline can be applied after each round of experimentation to build predictive models, extract key design rules, and generate data-driven design frameworks. I use this integrated, stepwise approach to optimize the structures, properties, and performance profiles of soft granular biomaterials for injection- and extrusion-based biomedical applications. In Part 3, I leverage the optimized materials from Part 2 to develop a novel microgel-based transcatheter additive manufacturing technology. I obtain proof-of-concept data for the platform's critical features, including controlled transcatheter material delivery to distant target locations, rapid in situ structuration of arbitrary 3D constructs, and reliable scaffold stabilization to ensure long-term implant integrity. Together, this work paves the way for minimally-invasive, patient-specific, in situ biofabrication.

Name thesis:degree_name

Doctoral

Department dc:contributor.department

Harvard-MIT Program in Health Sciences and Technology

Grantor dc:publisher

Massachusetts Institute of Technology

Year dc:date.issued

2023

Author dc:creator

• Verheyen, Connor Anthony

Advisor dc:contributor.advisor

• Roche, Ellen T.

Statement dc:rights

• In Copyright - Educational Use Permitted
• Copyright retained by author(s)

Licence dc:rights.uri

• https://rightsstatements.org/page/InC-EDU/1.0/

Handle dc:identifier.uri

https://hdl.handle.net/1721.1/151419

OAI identifier oai:identifier

oai:dspace.mit.edu:1721.1/151419