Abstract
The interaction between the central nervous system (CNS) and immune system is crucial in maintaining homeostasis. Scientists have recently realized the extreme significance of the role of CNS immune cells after injury, in aging, and in neurodegenerative disease. Morphological changes of microglia cells, which are immune cells in the brain, can reveal the state of the CNS. However, manual quantification of these complex morphologies is tedious, error-prone, and potentially biased. The primary objective of this thesis is to provide an automatic image-based engineering solution to study microglia structure and motion.
We propose a fully automatic system for quantifying 3D images of glial morphology over time to produce accurate image-based bioinformatics in naive and diseased settings. The quantification of morphology requires acquiring consistent digital reconstructions of the morphology, which is a challenging open problem in bioimage analysis. First, we describe an automatic 3D segmentation algorithm, called the coupled Tubularity flow field-Blob flow field (Tuff-Bff) for images of microglia. Tuff-Bff introduces a geometric deformable model designed to simultaneously reconstruct the large cell body and thin tubular processes. Our method found a 20% performance increase against state-of-the-art segmentation methods on a dataset of 3D images of microglia even in images with intensity heterogeneity throughout the object. The coupled Tuff-Bff segmentation results also yielded 40% improvement in accuracy for the ramification index of the processes, which reveals the efficacy of our method.
We also provide a methodology, called Hieroglyph, for consistent reconstruction of morphology over time using a novel hierarchical graph matching of glyphs, a term we use to describe the graph theoretic tree representation of glia. Our temporal graph representation possesses information about the connections between the paths of a cell and node in the path. This information is used to track the digital reconstruction at subsequent time frames. These temporal glyphs contain all the complex morphological data for a glia in space and time. Hieroglyph yielded a 21% performance increase compared to the state-of-the-art automatic skeleton reconstruction methods and outperforms the state of the art in different measures of consistency on datasets of 3D images of microglia.
To improve glial tracing we introduce C3VFC (Critical points on constrained Concentric Circles using Vector Field Convolution), which utilizes vector field convolution for detection and labeling of multiple cells in 3D images over time, leading to multi-object reconstruction. The C3VFC reconstruction results yielded more than 50% improvement on the next best performing tracing method. C3VFC achieved the highest accuracy scores, in relation to the baseline results, in four of the five different measures: entire structure average, the average bi-directional entire structure average, the different structure average, and the percentage of different structure.
Finally, we show that our automatic digital reconstruction system can provide a set of image-based bioinformatic measures for glia morphology and motility, including the volume covered, path length, path velocity, and bifurcation angle. We use the results from the three reconstruction algorithms to determine useful quantitative measurements to determine surveillance and ramification of microglia in naive, diseased, and injured animal models.
