University of Cambridge
Prediction of Coronary Atherosclerosis Development: A Study Based on the Combination of Mechanomics and Radiomics
Abstract
dc:description.abstractAtherosclerosis is the precursor to cardiovascular diseases (CVDs), the leading cause of death and disability globally. Coronary artery disease (CAD) from atherosclerosis significantly contributes to CVDs. Despite their primary role in delivering oxygenated blood to the myocardium, the anatomical development of coronary arteries and the pathological development of atherosclerosis within these arteries remain crucial research topics. Advancements in imaging hardware and techniques have established coronary computed tomography angiography (CTA) as the first-line non-invasive method for diagnosing CAD. Alongside, improvements in computational power have accelerated the growth in machine learning and significantly reduced the time for computational fluid dynamics (CFD) simulations. My PhD research leveraged high-throughput data from coronary CTA, including direct radiomic and morphological features, as well as indirect features from CFD simulations, to analyse correlations and develop predictive models for atherosclerosis development proximal to myocardial bridging (MB). To manage a relatively large dataset for coronary CFD, I developed a platform that ensures standardisation and repeatability and enables batch processing and automation. Throughout the development of this platform, bottlenecks in the coronary CFD workflow were identified and subsequently alleviated. Systematic analysis of distal branch effects on haemodynamics in the distal region of interest (ROI) revealed that while the influence of distal branches was indirect, maintaining accurate flow distribution at the ROI was crucial (Chapter 3). This insight led to the refinement of the flow distribution algorithm to maintain flow consistency at the ROI, even with the intentional removal of distal branches, thus enhancing work efficiency and reducing time and computational demands (Chapter 4). Additionally, given the rigidity of coronary geometry in CFD simulations, I conducted a comparative analysis of the most distinct diastolic and systolic cardiac phases to assess the importance of phase dependency (Chapter 5). This analysis revealed that phase dependency was not critical for the objectives of my study, thereby simplifying the simulation process. Moreover, my study involved applying this platform to a large dataset of patients with interarterial anomalous aortic origin of the coronary artery (ARCA) to explore how coronary morphology in two phases correlates with angina (Chapter 6). This effort identified a novel marker linking anginal symptoms with morphological features derived from diastolic and systolic phases. Ultimately, all previous work converged in applying the developed platform to MB patients, enabling me to discover correlations and make predictions using batch-generated radiomic, morphological, and mechanomic features (Chapter 7). This effort underscored the superior performance of radiomic features and the potential of multi-omic approaches. In summary, my PhD leveraged advancements in big data, machine learning, and computer-aided batch processing to extract reproducible and high-throughput features from coronary CTA images. This work has paved the way for novel markers and provided insights into the correlation between pathological atherosclerosis and physiological symptoms, particularly focusing on congenital anomalies in the coronary arteries.
Degree
thesis:*- Name dc:type.qualificationname
- Doctor of Philosophy (PhD)
- Level dc:type.qualificationlevel
- Doctoral
- Grantor dc:publisher.institution
- University of Cambridge
- Year dc:date.issued
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Zheng, Jin
- Advisor dc:contributor.advisor
-
- Graves, Martin
Subjects
dc:subject × 5Rights
dc:rightsIdentifiers
dc:identifier.*- DOI dc:identifier.doi
- https://doi.org/10.17863/CAM.113135
- OAI identifier oai:identifier
- oai:www.repository.cam.ac.uk:1810/375422