Identifying Key Gait Features Underlying Locomotor Deficits and Immunomodulatory Drug Therapy-Induced Recovery After Cervical Spinal Cord Injury in Rats (implementation of CLASSify)

From Dr. Aaron Silverstein, “I used CLASSify to analyze CatWalk® gait data from rats with spinal cord injuries, which included over 200 locomotor parameters. CLASSify helped identify the most important features using SHAP scores, allowing them to better understand how injury impacted movement. It was a powerful tool for simplifying complex data and highlighting key behavioral deficits.”

Dr. Silverstein wanted to utilize machine learning (ML) and AI techniques to help identify which of the many locomotor function variables measured by CatWalk® are most important for describing locomotor deficits and recovery following a C2 hemisection model spinal cord injury in rats. 16 variables of interest have been manually selected by their team, based upon their understanding of the injury model and the variables themselves. It would greatly enhance the rigor and efficiency of the analysis to have an objective, data-driven basis for choosing which variables are most important and useful to the field (i.e. CLASSify).

Two different manuscripts either have been, or will be submitted having to do with the data CLASSify helped with. One is currently under review and the other will be submitted for review. Those references are below.

Silverstein, A.L., Calulot, C.M., McLouth, C.J., Gensel, J.C., and Alilain, W. J. (2025). Experimentally induced C2 hemisection spinal cord injury induces locomotor deficits measurable by the CatWalk XT system in adult female rats. In review.

Silverstein, A.L., Calulot, C.M., McLouth, C.J., Gensel, J.C., and Alilain, W. J. (2025). Liposome-encapsulated clodronate and COX-2 inhibitor treatment impair ventilatory recovery but improve locomotor function following cervical spinal cord injury in rats. In preparation.

Supported by NIH R01 NS116068 to John Gensel and Warren Alilain.