A Deep Learning-Based Approach for Unbiased Kinematic Analysis in CNS Injury

Spinal cord injury (SCI) and multiple sclerosis (MS) are debilitating neurological conditions that impact millions worldwide. SCI leads to axonal damage, blood-spinal cord barrier disruption, and cell death, resulting in paralysis and sensorimotor deficits. MS is marked by oligodendrocyte death, demyelination, and immune infiltration, causing varied neurological impairments. While progress has been made in understanding these diseases, current assessment methods like the BBB, BMS, and EAE clinical scoring systems lack sensitivity, are prone to bias, and fail to detect subtle motor changes.

To overcome these limitations, I developed two markerless kinematic analysis systems, MotorBox and MotoRater, utilizing deep-learning algorithms from DeepLabCut. These tools were validated against traditional methods, demonstrating accurate, unbiased evaluations of movement. Using MotorBox, I detected significant motor changes in EAE mice previously considered asymptomatic, as well as transient motor impairments in SCI mice. The MotoRater revealed that SCI mice have worse motor outcomes in chronic stages, contrasting with BMS scores. Additionally, both systems detected anxiety-like behaviors in SCI and EAE models.

These findings suggest that the prodromal locomotor changes in EAE could predict outcomes, and the motor phenotypes in SCI could provide insights into disease progression. Integrating MotorBox and MotoRater with traditional scoring methods enhances accuracy, sensitivity, and reproducibility, providing valuable metrics for assessing therapeutic interventions and disease progression in SCI and MS models.