Image Analysis for Describing Meningeal Lymphatic Vessel Morphology

Neuroscientists have recently discovered the existence of meningeal lymphatic vessels in the brain and have shown their importance in preventing cognitive decline. With age, lymphatic vessels narrow, poorly draining cerebrospinal and interstitial fluids, which leads to plaque accumulation, a hallmark of Alzheimer’s disease. The analysis and detection of these vessels is performed by hand, and thus suffers from quantification variability. Furthermore, the only existing complexity measures currently extracted from images of these vessels are width and area, which are insufficient to capture morphological differences. This dissertation details the first automated segmentation and analysis methods developed for lymphatic vessels. The proposed segmentation approach, called LyMPhi, is a level set segmentation method featuring hierarchical matting to pre-determine foreground and background regions. The resultant approach eliminates the need for user-defined initialization, an advantage over competing methods, and produces smooth segmented contours. The level set force field is modulated by the foreground information computed by matting, while also constraining the segmentation contour to be smooth. Segmentation output from this method has a higher overall Dice coefficient and boundary F1-score compared to that of competing algorithms. The algorithms are tested on real and synthetic data generated by our novel shape deformation based approach. LyMPhi is also more stable under different initial conditions than comparative level set segmentation methods. Analysis can also be extended to studying elastic deformation for lymphangiogenesis as well as vessel narrowing. The deformation model can additionally be used for stretching existing vessel data into realistic synthetic data. Machine learning is explored and tested with application to segmenting the meningeal lymphatic vessels, showing promising results for the future. Manual segmentation, which is used as labels for neural network training, as well as a comparison for measuring segmentation accuracy, is analyzed with respect to statistical measures of variance. The last step in vessel analysis is to create relevant measures, many of which can be extended from already existing measures for other vessel networks. Given this work, neuroscientists have the first image analysis tools for meningeal lymphatic vessels, and subsequently can accurately capture information that will enable them to explore the vessels’ role in neurodegenerative diseases.