MARS_ML

MARS_ML performs automated pose estimation and deep learning-based classification of social interactions between pairs of mice to quantify behaviors such as grooming, play, and aggression for behavioral neuroscience research.

Key Features:

• Deep learning-based analysis: Employs computer vision and deep learning algorithms for feature extraction and behavior classification from video.
• Pose and trajectory extraction: Extracts poses and trajectories from videos of pairs of interacting mice for downstream analysis.
• Behavior classification: Classifies social interactions and specific behaviors such as grooming, play, and aggression.
• Novel feature extraction for accuracy: Uses novel feature extraction methods from poses and movements to improve classification accuracy.
• Real-time annotation: Performs real-time classification and annotation of recognized behaviors.

Scientific Applications:

• Behavioral neuroscience: Quantifies social interactions to investigate mechanisms underlying social behavior.
• Genetics and developmental biology: Assesses the effects of genetic modifications and developmental changes on mouse social behavior.
• Neurological disorder research: Tracks behavioral phenotypes and progression relevant to neurological disorders.
• Large-scale behavioral studies: Enables automated analysis to scale studies that are impractical with manual observation.

Methodology:

Video footage of interacting mice is captured and processed through deep learning algorithms trained to recognize specific postures and movements corresponding to behaviors (e.g., grooming, play, aggression), and the system classifies these behaviors in real-time to provide detailed behavioral data.