Honors Thesis: Unveiling Tissue-Level Biomarkers in Contracted Rat Elbows Using Deep Learning

Abstract: Post-traumatic joint contracture in the elbow arises from traumatic soft-tissue injuries, dislocations, or fractures, leading to restricted range of motion and stiffness. Current qualitative morphological characterization of this condition hampers quantitative understanding and effective treatment planning. Leveraging non-invasive MRI (Magnetic Resonance Imaging) scans, machine learning techniques offer promise in extracting tissue-level biomarkers for deeper comprehension and treatment strategies. In this study, we explore the application of deep learning, particularly in three key areas: super-resolution reconstruction of low-resolution MRI scans, assisted region of interest tracing using deep segmentation, and tissue type identification through unsupervised semantic segmentation. Our results demonstrate the feasibility of these approaches, showcasing advancements in imaging resolution, segmentation accuracy, and tissue classification, thus providing valuable insights for clinical diagnosis and treatment planning in post-traumatic joint contracture.

Advisor: Dr. Ulugbek Kamilov