Automated extraction of blood vessel networks from 3D microscopy image stacks via multi-scale principal curve tracing

Blood vessel segmentation, that is, extraction of the center lines and corresponding local cylinder radii are important for the study of vascular diseases, and in the brain also important for the modeling and understanding of relationships between hemodynamics and electrical neural activity. Several image processing methods have been proposed for vessel extraction in many domains including those that explore the use of pattern recognition techniques, model-based approaches, tracking based approaches, artificial based approaches, neural network based approaches, and miscellaneous tube-like object detection approaches. In this paper, we propose a ridge tracing approach based on recently developed principal curve (PC) projection and tracing algorithms for the extraction of vasculature networks in the brain from 3D microscopy image stacks. Results on mice brain imagery obtained for the purpose of studying hemodynamic effects on neural activity are promising.