A statistical framework for estimation of cell migration velocity

Migration velocity of cell populations in vitro is one of important measurements of cell behaviors. As there are massive amount of cells in one image that share similar characteristics and are highly deformable, it is often computational expensive to track every individual cell. It is also difficult to track cells over a long period of time due to propagation of segmentation and tracking errors. This paper presents an algorithm to estimate migration velocity of cell populations observed by time-lapse microscopy. Instead of tracking cells individually, our proposed algorithm computes mutual information between image blocks of consecutive frames. The migration velocity is then estimated by a linear regression, with mutual information and foreground area ratio as input. Experiments on a variety of image sequences verified that our algorithm can give accurate and robust estimation under different situations in real-time