Cancer Cell Tracking
Identifying cell trajectories is an important step in analyzing physiological events in computerized Video Time-lapse Microcopy. The large variety and transformation of cell shapes and cells' Brownian motion make cell tracking a challenge problem. In this paper we present a cell tracking system, implemented as a particle filter within texture-adaptive active contour formulations. The texture-adaptive weights on the external energy of the active contour model enables snakes to bypass internal psuedo-edges and stop on low-contrast cell boundaries. Using the texture of cells as observation model, we can track cells whose locations follow a multimodal distribution with a particle filter. This system is a novel combination of tracking algorithms and deformable models, and allows for the first time to automatically track non-fluorescence cellular microscopy images. The implemented tracker is tested on both normal and autophagy cell image sequences, to demonstrate the slow motion properties of cells in autophagy.