Behavioral Analysis of Test Subjects Using Pose Detection

Abstract

This project seeks to design a machine learning behavioral analysis framework to discover the underlying mechanisms of neurodevelopmental disorders and the effectiveness of the treatment. Unique action patterns are hallmarks of neurodevelopmental disorders which could be caused by Traumatic Brain Injury (TBI), and vary considerably from individual to individual. In animal models, conventional behavioral phenotyping captures limited fine-scale variations. This project aims to design a deep learning model to characterize mouse movement on multiple timescales using the open-field test. The framework takes virtual markers from pose estimation to find behavior clusters and generate signatures of behavior classes. This tool measures spatial and temporal habituation to a new environment across minutes and days, different types of selfgrooming, locomotion and gait. It also tests under task or social conditions which would reveal more information about behavioral dynamics and variability.