Laboratory Rat Tracking Methods in Biomedical Experiments

Presented are two video analysis methods for tracking laboratory rats in three biomedical scenarios. In the first scenario, rat moves freely in a rectangular box, necessitating computation of the number of visits and visit duration in specific box areas. To assess the recognition of novel objects in the second scenario two toys are placed on the diagonal of the box, requiring tracking of visits and visit durations for each toy. The third scenario involves a Y-maze environment, with calculation of visits and visit durations for each branch. Initially, a tracking method using a voting procedure was proposed for the first scenario. The method combines results from binarized body detection and identification of the main cluster in the difference image of two consecutive frames. Subsequently, a deep learning-based body localization technique was applied across all three scenarios.