Rat Tongue Kinematics with Markerless XROMM Using DeepLabCut and Marker Masking.

Area of Science:

• Biomechanics
• Animal Behavior
• Biomedical Imaging

Background:

• X-ray Reconstruction of Moving Morphology (XROMM) is crucial for 3D biomechanical studies in animals.
• DeepLabCut (DLC) offers markerless pose estimation but traditional XROMM methods struggle with missing markers.

Purpose of the Study:

• To develop a robust XROMM workflow using DLC for improved 3D marker tracking.
• To enable accurate pose estimation even with incomplete or absent markers.
• To gain novel insights into rodent feeding behavior through tongue kinematics.

Main Methods:

• Utilized XMALab results for camera calibration to predict DLC-guided 3D XROMM markers.
• Implemented a marker masking augmentation for tracking with incomplete marker sets.
• Applied the method to extract tongue kinematic patterns in rodents.

Main Results:

• Demonstrated accurate XROMM pose estimation with subsets of visible beads or no visible markers.
• Showed improved robustness against missing markers compared to traditional methods.
• Successfully extracted tongue kinematic patterns, revealing novel aspects of rodent feeding behavior.

Conclusions:

• The proposed DLC-guided XROMM method enhances accuracy and robustness in biomechanical tracking.
• This approach allows for reliable pose estimation even with marker loss.
• The study provides new insights into rodent tongue biomechanics during natural feeding.