Human Spinal Epidural Neuromodulation Modeling and Stimulation Effect Prediction

Spinal neuromodulation via epidural electrical stimulation is a promising therapy for motor function restoration of patients with spinal cord injury. By applying specific sequences of electrical stimulation to the spinal cord through the electrode array, the patient can achieve multiple movements. However, one of main difficulties of this therapy is that the mapping between stimulation parameters and body motions is complex and ambiguous. To achieve specific actions, it often requires a large number of experiments to tune stimulation parameters. Individual differences of functional innervation and huge stimulation parameter space make the parameter-tuning process more challenging. It is also important to prevent potential damage to the patient while exploring new parameters. To better understand human spinal neuromodulation process and predict the effect of epidural electrical stimulation, we developed a personalized hybrid model with finite element method and biophysical neural computation. The high-precision finite element model of the spinal cord and electrode array was build based on CT and MRI of a spinal cord injured patient. Biophysical axon models were embedded in dorsal column, afferent and efferent nerve roots to calculate neural activities. The neural activation matrix was imported into a neural network to obtain the normalized activity intensity of targeted muscles. Learning from a small amount of stimulating parameters together with surface electromyography (EMG) data of the patient, our model can predict muscle activities of new stimulation parameters with high accuracy rate. The hybrid model can give reliable predictions whether a specific muscle would be strongly activated or not activated, which is very important for filtering out unsafe or ineffective parameters. The model could serve as a safe agent which computationally evaluate and predict effective stimulation parameters for motor function restoration. The model can also help with the optimization of the electrode array and revealing the mechanism of epidural spinal neuromodulation.