A deep learning-based approach for unbiased kinematic analysis in CNS injury

Area of Science:

• Neuroscience
• Biomedical Engineering
• Animal Models

Background:

• Traumatic spinal cord injury (SCI) affects over 300,000 individuals in the US, causing sensorimotor deficits and paralysis.
• Current therapeutic development for SCI is hindered by a lack of sensitive and reproducible functional assessment methods.
• Traditional locomotion tests like the Basso Mouse Scale (BMS) have limitations in objectivity and the range of measurable metrics.

Approach:

• Developed two novel marker-less kinematic analysis systems, MotorBox and MotoRater, utilizing deep-learning algorithms (DeepLabCut).
• MotorBox employs a custom-designed chamber, while MotoRater is integrated with a commercial device.
• Validated these systems against the established BMS test to assess their accuracy and sensitivity in detecting SCI-induced locomotor changes.

Key Points:

• MotorBox and MotoRater provide highly sensitive, accurate, and unbiased analysis of mouse locomotor behavior.
• These systems can detect subtle SCI-induced changes in movement metrics such as range of motion, gait, coordination, and speed.
• They enable the measurement of movement metrics not quantifiable by the BMS test, offering a more comprehensive assessment.

Conclusions:

• MotorBox and MotoRater offer robust, sensitive, and reproducible means of functional assessment for SCI research.
• Integration of these kinematic analyses with traditional scoring methods (e.g., BMS) enhances the rigor and breadth of behavioral outcome evaluation.
• These advanced tools are crucial for improving the success rate of clinical translation for SCI therapies.