Quantifying pathological changes to myelin with high resolution birefringence microscopy and deep learning

OA Version
Citation
Abstract
In the central nervous system (CNS), myelin, the insulating sheath around axons, enables rapid communication between brain regions, coordinating complex tasks such as cognition, memory, and motor function. However, insults to myelin due to neurodegeneration, aging, or injury lead to functional deficits, the underlying mechanisms of which remain incompletely understood. Accurate assessment of myelin integrity across large brain regions is essential for understanding disease progression and evaluating potential therapeutics. Electron microscopy and various optical imaging techniques have been demonstrated for high-resolution imaging of myelin pathology in post-mortem brain tissue, but lack the necessary imaging throughput and quantitative analysis required for true structural imaging of myelin. There remains a critical need for an imaging technique that combines high-resolution and high-throughput analysis of myelin, providing deeper insight into the impact of myelin degradation in different contexts. Birefringence microscopy (BRM) is an emerging technology which enables widefield (camera-based), label-free imaging for structural imaging of individual myelinated axons for analysis of myelin pathology. In this thesis, I present technical advancements that optimize BRM for high-throughput and multiscale structural imaging of myelin. I develop new imaging and analysis methods to enhance the quantitative capabilities of BRM, and identify key sample preparation protocols required for reliable, high-resolution imaging of individual myelinated axons. These optimizations establish BRM as a scalable tool for analysis of myelin integrity. Building on this, I validate BRM for identifying myelin pathology in two rhesus monkey models of myelin damage: (1) circumscribed cortical injury and (2) age-related cognitive decline. I also implement deep learning tools to aid in the automated quantification of myelin pathology across large brain regions. In both models, BRM enables robust detection of myelin defects and their spatial distribution, which are compared to behavioral and functional metrics. Finally, in the cortical injury model, I evaluate the efficacy of treatment with mesenchymal stem cell-derived extracellular vesicles (MSC-EVs) for promoting myelin repair, demonstrating BRM’s ability to detect treatment-driven changes in myelin structure. Together, these investigations highlight BRM as a powerful tool for scalable, high-resolution analysis of myelin damage and repair, with broad applicability to studies of neurodegeneration, injury, and aging.
Description
2025
License