Automated Rats Detection and Tracking for Behavioral Analysis in Biological Experiments

Abstract

The social behavior of rats is commonly employed as an initial model to explore the mechanisms behind various neurological illnesses in humans. Neuroscientists need a precise quantitative measure for behavior monitoring and real-time tracking in lab settings in order to investigate the relationship between brain systems and social behaviors. This work created a real-time detection and tracking method for mice in a lab setting, concentrating on the Morris Water Maze, a popular experimental setup for researching spatial learning and memory in rodents, in order to get around current real-time tracking limitations. Unlabeled recordings from the Morris Water Maze trials make up the dataset we used. After labeling the dataset and creating foreground masks using the background subtraction method, we utilized contours to separate the largest foreground contour-which represented the rat's body-from the rest of the dataset. Moreover, we employed the Kalman filter to forecast the rat's position in consecutive frames and YOLOv11 (You Only Look Once, v11) for detection. The model's recall was 74.14/% and its precision was 81.79%. The mAP averaged over thresholds from 0.50 to 0.95 (mAP@0.50-0.95) was 68.70%, whereas the mean average precision at an IoU threshold of 0.50 (mAP@0.50) was 83.60%, suggesting great detection accuracy. These findings show how well the model tracks mouse behavior in the Morris Water Maze, providing neuroscientists with a trustworthy instrument for analyzing spatial navigation and learning activities in real time. This method offers high-accuracy monitoring in dynamic lab environments, which has the potential to advance behavioral analysis in neurological research.