Tracking multiple mice

Abstract

Monitoring mouse social behaviors over long periods of time is essential for neurobehavioral analysis of social mouse phenotypes. Currently, the primary method of social behavioral plienotyping utilizes human labelers, which is slow and costly. In order to achieve the high throughput desired for scientific studies, social behavioral phenotyping must be automated. The problem of automation can be divided into two tasks; tracking and phenotyping. First, individual body parts of mice must be accurately tracked. This is achieved using shape context descriptors to obtain precise point to point correspondences between templates and mice in any frame of a video. This method provides for greater precision and accuracy than current state of the art techniques. We propose a means by which this tracking information can be used to classify social behaviors between mice.