An Active Learning Based Prediction of Epidural Stimulation Outcome in Spinal Cord Injury Patients Using Dynamic Sample Weighting

Recent studies suggest that epidural stimulation of the spinal cord could increase the motor pattern both in motor and sensory complete spinal cord injury (SCI) patients. However, choosing the optimal epidural stimulation variables, such as the frequency, intensity, and location of the stimulation, significantly affects maximal motor functionality. This paper presents a novel technique using machine learning methods to predict the functionality of a SCI patient after epidural stimulation. Additionally, we suggest a committee-based active learning method to reduce the number of clinical experiments required through exploring the stimulation configuration space more efficiently. This paper also introduces a novel method to dynamically weight the results of different experiments based on neural networks to create an optimal estimate of the quantity of interest. The proposed method for the prediction of stimulation outcomes is evaluated based on various accuracy measures such as mean absolute error, standard deviation, and correlation coefficient. The results show that the proposed method can be used to reliably predict the outcome of epidural stimulation on maximum voluntary contraction force with the prediction error of about 15%.