Machine Learning Algorithm to Predict Motor Improvement Induced by Epidural Stimulation After Spinal Cord Injury

Introduction: Spinal cord injury (SCI) leads to severe motor impairments, with limited treatment options for recovery. Epidural electrical stimulation (EES) has shown potential for rehabilitation, but predicting individual patient outcomes remains a challenge. This phase II trial evaluates motor recovery in paraplegic patients (ASIA A/B) following EES, with a focus on machine learning models to predict motor improvements.

Methods: Five paraplegic patients were implanted with EES devices and followed for 12 months. Outcomes were assessed using the Fugl-Meyer Assessment (FMA) for motor function, Berg Balance Scale (BBS), Ashworth Scale for spasticity, Beck Depression Inventory (BDI), WHO Quality of Life (WHO-QoL), Neurogenic Bladder Symptom Score (NBSS), and Neurogenic Bowel Dysfunction Score (NBD). Generalized linear methods were applied to identify the best predictor of motor improvement, and data were analyzed with MATLAB and JMP software.

Results: The Fugl-Meyer motor score increased by 78% over 12 months, with significant gains in the first 5 months (p < 0.0001). Balance (BBS, p < 0.001) and spasticity (Ashworth Scale, p < 0.05) also improved. The Gompertz 3-parameter (3P) model provided the best fit for motor recovery predictions, with a plateau observed at 5 months. Other scales, including BDI, WHO-QoL, NBSS, and NBD, showed trends toward improvement but were not statistically significant. Sensory recovery and autonomic functions remained stable.

Conclusion : EES leads to significant motor improvement in paraplegic patients, particularly in the early stages of treatment. Machine learning models successfully predicted motor recovery trajectories, providing insights for optimizing rehabilitation. These findings support the use of predictive analytics to tailor EES interventions and improve patient outcomes.