Computer Vision for Mouse Behavioral Phenotyping

Charles Frogner, Massachusetts Institute of Technology

Characterization of mouse behavior is central to the pursuit of mechanisms underlying numerous psychiatric disorders. We are developing a set of tools and techniques based on machine learning and computer vision to automate existing modes of behavioral phenotyping and establish new, quantitative metrics that can capture richer representations of behavior. We have developed a trainable computer vision system that can track multiple mice throughout their social interactions and categorize a variety of behaviors. A particular goal of this work is the characterization of social behaviors that model deficits associated with autism.

Abstract Author(s): Charles Frogner, Chun-Kai Wang, Stav Braun, Nick Edelman, Hueihan Jhuang, Yuzhao Ni, Wesam Manassra, Guoping Feng, Tomaso Poggio