A team from the Life Sciences Institute at the University of Michigan has developed a new software tool to help life scientists analyze animal behavior more effectively. Open-source software LabGym uses AI and deep learning to identify, categorize, and count behaviors defined in various animal model systems.
Scientists measure animal behavior for a variety of reasons, such as to better understand how a particular drug may affect an organism or to map how circuits in the brain communicate to produce a particular behavior.
Researchers in the lab of University of Michigan faculty member Bing Ye, Ph.D., are studying neural development and the defects that lead to brain disorders. They analyze the movements and behavior of Drosophila melanogaster (fruit flies) as a model to study the development and functions of the nervous system. Because fruit flies and humans share many genes, their studies of fruit flies often provide insight into human health and disease.
Yujia Hu, Ph.D., a neuroscientist in Bing Ye’s lab at the University of Michigan Life Sciences Institute and lead author of the study describing LabGym, the new software, explains:
“Behaviour is a function of the brain. Therefore, analyzing animal behavior provides essential information about brain function and how it changes in response to disease.”
However, manually identifying and counting animal behavior is time-consuming and, on the other hand, very subjective for the researcher analyzing the behavior. Existing software for automatically quantifying animal behavior, meanwhile, presents challenges.
Bing Ye, who is also a professor of cell and developmental biology at the medical school, comments:
“Many of these behavior analysis programs are based on predefined definitions of a behavior. For example, when a Drosophila larva rotates 360 degrees, some programs count a roll. But why doesn’t 270 degrees also matter? Many programs don’t necessarily have the flexibility to count this without the user knowing how to transcode the program..
Replication of the human cognitive process
To overcome these challenges, Yujia Hu and his colleagues set out to develop a new program that better mimics the process of human cognition, “thinks” more like a scientist, and is easier to use for biologists without programming skills.
Researchers provide examples of the behavior they want to analyze. LabGym then uses deep learning to improve its ability to recognize and quantify behavior.
While other research, such as that presented by ETH Zurich last June, allows animal behavior to be analyzed using a single video, the program combines video data with model images.
To help the program recognize behaviors, Yujia Hu created a still image showing the animal’s movement pattern by merging the outlines of the animal’s position at different times. The team found that combining the video data with the sample images increased the program’s accuracy and cognitive flexibility.
Typically, when researchers manually analyze behaviors in a video, they ignore irrelevant information (such as a static background).
LabGym is designed to ignore this background information and consider both the animal’s overall movement and changes in position in space and time, as a human researcher would, and allows for the simultaneous tracking of multiple animals.
Another key feature of LabGym, according to Bing Ye, is its biodiversity because although it was developed with Drosophila, it is not limited to a single species.
He comments:
“Actually, it’s rare. It was written for biologists to adapt to the species and behavior they want to study without the need for programming skills or supercomputers.”
Pharmacologist Carrie Ferrario, Ph.D., offered Bing Ye and her team to test and refine the program in the rodent model system she works with.
As Associate Professor of Pharmacology and Associate Associate Professor of Psychology, she studies the neural mechanisms that contribute to addiction and obesity using rats as a model system. To complete the necessary observation of drug-induced behavior in animals, she and members of her laboratory team relied largely on manual scoring, which, as we have seen, is subjective and takes a long time.
She explains :
“I’ve been trying to solve this problem since grad school, but the technology in terms of artificial intelligence, deep learning, and computers just wasn’t there. This program solved an existing problem for me, but it is also very useful. I see the potential that it can be useful for analyzing animal behavior under almost unlimited conditions.”
The team then plans to further refine the program to improve its performance in even more complex conditions, such as observing animals in the wild.
Sources of the article:
Department of Biological Sciences, University of Michigan.
Learn :“LabGym: Quantifying Custom Animal Behaviors 1 with Holistic Learning-Based Assessment“,
DOI: 10.1016/j.crmeth.2023.100415
authors : Yujia Hu, Carrie R. Ferrario, Alexander D. Maitland, Rita B. Ionides, Anjesh Ghimire, Brendon Watson, Kenichi Iwasaki, Hope White, Yitao Xi and Bing Ye, University of Michigan, and Jie Zhou, University of Northern ‘Illinois .
The open source code and a complete LabGym manual can be found on the GitHub page: https://github.com/umyelab/LabGym