Segmentation, Analysis and Visualization of Large Microvascular Networks in the Mouse Brain

Advances in high-throughput microscopy allow researchers to collect three-dimensional images of whole organ vascular networks at sub-micrometer resolution. Microvascular networks are highly sparse and space-filling, making them difficult to segment, visualize, and analyze. Since microvasculature plays a prominent role in tissue function and development, overcoming these limitations could lead to advances in disease diagnosis and grading. There is therefore a compelling need for algorithms that segment, visualize, and analyze organ-scale microvascular networks at sub-cellular resolution. In this dissertation, I present a platform enabling scalable segmentation, visualization, and analysis of microvascular networks encoded within multi-terabyte three dimensional microscope images. I first propose a highly parallel and scalable algorithm to segment microvascular networks on GPUs, providing domain experts with an accessible tool usable on standard workstations. I then propose multiple visualization methods that enable the study of complex microvascular networks, highlighting important structural features that are challenging to evaluate with traditional volumetric and rasterization tools. Finally, I demonstrate that my platform can be extended to collect statistical features and identify metrics for tissue classification and characterization.