Microvasculature network identification in 3-D fluorescent microscopy images

Segmentation of noisy and low-resolution images of microvasculature from 3-D fluorescence microscopy has been proven a challenging task. In this paper, we propose an approach to identify the global connectivity structure in a microvasculature network, which can be of use in obtaining more detailed segmentation results and for comparing and validating other network segmentation methods. Our approach begins with determining a collection of potential vessel junctions. Thinking of the junctions as vertices in a graph modeling connectivity of the vasculature network, we formulate and solve an integer linear programming problem based on a novel image-derived utility function to remove false junctions from the previously identified set and determine the network structure of remaining junctions. The scheme is demonstrated to be robust and computationally tractable making it well suited for use in high-throughput applications. Quantitative validations demonstrating high rate of sensitivity are given.