MonkeyTrail: A scalable video-based method for tracking macaque movement trajectory in daily living cages

Area of Science:

• Neuroscience
• Animal Behavior
• Computational Biology

Background:

• Behavioral analysis of macaques is crucial for neuroscience research.
• Existing automatic animal behavior analysis methods struggle with occlusion and environmental changes in daily living cages.
• Accurate tracking of macaque movement trajectories in naturalistic settings is underdeveloped.

Purpose of the Study:

• To introduce MonkeyTrail, a novel method for extracting and analyzing daily movement trajectories of macaques in their living cages.
• To overcome limitations of previous methods by enabling analysis in standard environments without specific setup.
• To provide a low-cost, scalable solution for long-term behavioral monitoring.

Main Methods:

• MonkeyTrail utilizes frequently generated virtual empty backgrounds combined with background subtraction to isolate macaque subjects.
• The virtual background generation integrates the frame difference method (FDM) and the YOLOv5 deep learning model.
• Performance was validated against ground-truth data from over 8,000 labeled video frames.

Main Results:

• MonkeyTrail demonstrated superior tracking accuracy and stability compared to YOLOv5, Single-Shot MultiBox Detector, frame difference method, and naive background subtraction.
• The method successfully analyzed long-term surveillance videos to assess changes in macaque behavior, including movement amount and spatial preference.
• Low-cost hardware implementation was confirmed, suitable for individually caged macaques.

Conclusions:

• MonkeyTrail offers a robust and accurate solution for automated behavioral analysis of macaques in their daily environments.
• The method facilitates low-cost, large-scale studies, advancing neuroscience research through detailed behavioral insights.
• This approach addresses a critical gap in the analysis of primate behavior in ecologically relevant settings.