KineWheel-DeepLabCut Automated Paw Annotation Using Alternating Stroboscopic UV and White Light Illumination.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Vision

Background:

• Rodent pose tracking is crucial for understanding neural circuits, behavior, and dysfunction.
• Markerless pose estimation tools like DeepLabCut require significant manual annotation for neural network training.
• Reducing human labor in annotation can accelerate neuroscience research.

Purpose of the Study:

• To develop an automated method for generating annotated rodent paw placement datasets.
• To reduce the human effort required for training neural networks used in rodent pose tracking.
• To improve the efficiency of markerless pose estimation in neuroscience.

Main Methods:

• Utilized invisible fluorescent markers activated by UV light.
• Employed stroboscopic alternating illumination (UV and white light) at 720 Hz.
• Developed a color filtering and paw location mapping technique.
• Transferred paw annotations from UV frames to white light frames for neural network training.

Main Results:

• Successfully demonstrated an automated approach for rodent paw placement annotation.
• Achieved high-quality position annotations for markerless tracking of mouse paws.
• Significantly reduced the need for manual human annotation in dataset creation.
• Validated the method using a KineWheel-DeepLabCut setup.

Conclusions:

• The developed automated method streamlines neural network training for rodent pose tracking.
• This open-source approach enhances efficiency and reduces labor in neuroscience research.
• Facilitates more accessible and widespread use of advanced pose estimation techniques.