An Automated Tracking System for Y-Maze Behavioral Test using Kinect Depth Imaging

This paper presents an image processing system for automated tracking and behavior analysis of the popular Y-maze test on freely behaving rats, using depth imaging provided by a Microsoft Kinect (R) 2D/3D imager. A contour-based segmentation algorithm was developed to identify the maze shape and extract its arm and center divisions. Using the extraction results, the system is capable of tracking the animal position and arm entry sequence for calculating spontaneous alternations and other measures that are used in analyzing the animal working memory and activity. The system was validated in vivo on seven freely behaving rats, and the results showed perfect agreement with human annotations, 100% accuracy in arm entry tracking and less than 0.1 s error in time stamps of "enter/leave" actions.