DGPC-Net: Dual Branch Gaussian Process Constrained Network for Spinal Cord Stimulation Effect Prediction

This paper presents DGPC-Net, a novel dual-branch Gaussian Process Constrained Network designed to predict muscle activation intensity resulting from spinal cord stimulation (SCS) across various electrode configurations and stimulation amplitudes. To evaluate the performance of the model, real world MRI scan, finite element simulation and NEURON simulation were used, creating a personalized virtual environment to generate simulated data of muscle activation in response to epidural electrical stimulation (EES). The results demonstrate that DGPC-Net achieves high predictive accuracy of muscle activations, with a balanced accuracy of 81.42% in binary classification (activated or not activated) and a strong correlation of 0.93 between predicted and ground-truth muscle activation intensities. Additionally, ablation studies confirm the necessity of each component within the network. The model's ability to predict muscle activation in real time offers the potential for reducing the time required for parameter optimization, and improving the efficiency of treatments for conditions such as spinal cord injury and gait issues in Parkinson's disease.Clinical relevance- This research presents DGPC-Net, a novel tool for predicting muscle activation intensity in response to SCS. This approach has the potential to significantly re-duce the time and effort required for parameter optimization, enhancing treatment efficiency and outcomes.