Automated Monitoring of Animal Behavior in Medical Research

Project: Smart Vivarium
PI: Serge Belongie, Prof, Jacobs School of Engineering
Funding: Calit2
Division: UCSD
Corporate Partner: Ericsson Wireless
Start Date: January 2003

In a pilot project funded exclusively so far by Calit2, computer scientists and animal care experts at UCSD have come up with a new way to automate the monitoring of mice and other animals in laboratory research. Combining cameras and distributed/embedded sensors with elements of computer vision, information technology and artificial intelligence, this project aims to maximize the quality of animal research, while at the same time enabling better health care for animals. "Today a lot of medical research relies on drug administration and careful monitoring of large numbers of live mice and other animals, usually in cages located in a vivarium," says Serge Belongie, principal investigator on the Smart Vivarium project. "But it is an entirely manual process, so there are limitations on how often observations can be made, and how thoroughly those observations can be analyzed."

Belongie put together an interdisciplinary team to develop the hardware and software for automated, 24-hour-a-day monitoring and archiving of a continuous stream of measurements on animal behavior - rather than periodic observations by a lab technician. So far, Belongie has demonstrated the computer-vision and pattern-recognition software with data from a single cage, but the plan is to embark on a multi-million-dollar project that would allow researchers to develop and deploy the technology for two key areas - medical research, and emergency response (for so-called 'sentinel' cages at critical facilities).

UCSD is a major biological sciences research center, and animal-care specialists believe the technology under development could dramatically improve the care of research animals. The Smart Vivarium could make better use of fewer lab animals and lead to more efficient animal health care, while sick animals could be detected and diagnosed sooner, allowing for earlier treatment. The technology would also help to reduce the number of animals needed in scientific investigations, because by recording all the data the first time, scientists can go back and look for different patterns in the data without using more mice to perform new versions of the experiment. For many of the same reasons, the underlying technology could be useful for the early diagnosis and monitoring of sick animals in zoos, veterinary offices and agriculture.

Contact: Serge Belongie, (858) 822-5163, sjb@cs.ucsd.edu