Yi Jung Cheng
Research Mentor(s): Bing Ye
Mentor Department: Life Sciences Institute, University of Michigan
Authors:
Session: Session 1 (9:00am – 9:50am)
Presentation Type: Poster 94
Abstract
Quantifying animal behavior is crucial for biological research, but current computational tools rely on high-level properties like body poses, limiting holistic assessment. LabGym addresses this gap by providing a more comprehensive approach to behavioral analysis, particularly for complex behaviors such as social interactions and subtle changes in behavior. LabGym utilizes advanced machine learning techniques, including deep neural networks, to analyze animal behavior. The process involves three main steps: 1) Preprocessing video footage to optimize quality and reduce noise. 2) Training LabGym on user-defined behaviors, allowing for customization across various species and behavior types. 3) Applying the trained model to quantify and analyze specific behaviors of interest. In collaboration with researchers Drs. Ling-Yu Liu and Joshua Emrick in the School of Dentistry, we aim to measure changes in the breathing patterns of anesthetized mice before and after light pulse stimulation. LabGym’s ability to capture subtle behavioral changes makes it ideal for this application, potentially revealing nuanced alterations in respiratory function that may not be detectable through traditional methods. Its application in studying anesthetized mice’s breathing patterns could provide valuable insights into the effects of light stimulation on respiratory function, potentially informing anesthesia protocols and neuroscience research. LabGym’s versatility in capturing multi-individual interactions and non-social behaviors makes it a powerful tool for a wide range of biological studies, from basic science to drug development. By enabling more accurate and comprehensive behavioral analysis, LabGym has the potential to bridge the gap between neural recording techniques and behavioral outputs, advancing our understanding of brain-behavior relationships.
