Machine learning based computational framework for medical imaging studies

Abstract

Achilles tendinopathy is a common and stubborn injury, particularly among runners and athletes. It causes pain, restricts movement, and can be frustratingly difficult to manage long term. While conventional imaging tools like ultrasound and standard MRI can aid in diagnosis, they often miss the subtle structural changes that matter most—especially during recovery.
Given these limitations, this study was motivated by the need to explore a computational approach that combines advanced imaging—specifically Ultra-Short Echo Time MRI (UTE-MRI)—with Machine Learning to better track tendon health. Using techniques like histogram and cluster analysis, the framework identifies meaningful patterns in tissue composition that correspond to both degeneration and healing phases. A randomized clinical trial involving collagen peptide supplementation and eccentric exercise provided real-world data to test and refine the method.
To streamline analysis and reduce manual effort, we also developed a deep learning model for automatic tendon segmentation. The model showed reliable performance across different datasets, showing potential for routine clinical use.
Overall, the findings suggest that pairing UTE-MRI with smart computational tools can offer a clearer, more objective view of tendon changes over time. This approach may help clinicians detect tendinopathy earlier, personalize treatment strategies, and monitor rehabilitation more precisely—ideally improving outcomes for patients with this often-persistent injury.

Date of Award	2026
Original language	English
Supervisor	Girija CHETTY (Supervisor) & Roland GOECKE (Supervisor)