Analyzing Leukocyte Migration Trajectories by Deformable Image Matching

This paper presents a method for analyzing leukocyte migration in live animals by image processing. For detecting leukocyte movements, we adopt a tracking-by-matching approach that compares between consecutive time-frames in a time-lapse motion image observed with intravital microscopy. Since leukocytes migrate at a large variety of speeds, we developed a tool based on a matching algorithm, called DeepMatching, which uses a deep, multi-layer, convolutional architecture designed for matching images and can therefore efficiently determine dense correspondences in the presence of significant changes between images. DeepMatchihg was originally designed in computer vision research, and we modified the code to apply for cell tracking. Also we extended the matching method to handle not only 2D movements but also depth-directional ones in 3D motion images. By linking the leukocyte movements into their trajectories, we have detected leukocyte gathering points that are estimated as high concentration points of chemo-attractants. It is suggested that leukocytes may act as living micro-probes to the concentration of allergens in contact dermatitis by our method.