Automated Behavioral Analysis System

­Overview

The manual annotation of video data in the life sciences creates a number of problems, including those resulting from inherent biases. We have developed a neural network system to identify and automate the labeling of behavior in videos of animal test subjects.

Market Opportunity

Many areas of the life sciences demand the manual annotation of large amounts of video data. Not only does this create major bottlenecks, but a number of controversies in behavioral studies have arisen because of the inherent biases and challenges associated with the manual annotation of behavior. Many of these issues could be resolved through the use of objective computerized quantitative techniques.

Innovation and Meaningful Advantages

We have developed a high-throughput system for the automated monitoring and analysis of rodent behavior. Our approach capitalizes on recent developments in deep learning, a branch of machine learning that enables neural networks composed of multiple processing stages to learn visual representations with multiple levels of abstraction. Our system accurately recognizes normal and abnormal rodent behaviors at a level indistinguishable from human scoring of the typical behaviors of a singly housed mouse from video. We have also developed convolutional and recurrent-convolutional neural network architectures to process data.

Our system leverages machine learning and computer vision to analyze large volumes of data and discover novel visual features of behavior that are hidden to the naked eye. Once trained, the neural network is able to identify observed behaviors of test subjects in a video and annotate the time the observed behavior occurred, creating an annotated video that can be stored for future reference.

Collaboration Opportunity

We are interested in exploring 1) startup opportunities with investors; 2) research collaborations with leading companies; and 3) licensing opportunities with companies.

IP Information

Patent US 10,181,082 B2, Issued January 15, 2019

Publications

Jhuang H,  Garrote E, Yu X, Khilnani V, Poggio T, Steele AD, Serre T. Automated home-cage behavioral phenotyping of mice. Nature Communications 2010 Sept 07;68(2010). doi.org/10.1038/ncomms1064.