Toward a generalizable deep CNN for neural drive estimation across muscles and participants

Abstract

High-density electromyography (HD-EMG) decomposition algorithms are used to identify individual motor unit spike trains, which collectively constitute the neural code of movements, to predict motor intent. This approach has advanced from offline to online decomposition, from isometric to dynamic contractions, leading to a wide range of neural-machine interface applications. However, current online methods need offline retraining when applied to the same muscle on a different day or to a different person, which limits their applications in a real-time neural-machine interface. We proposed a deep convolutional neural network (CNN) framework for neural drive estimation, which captures general spatiotemporal properties of motor unit action potentials to generalize its application without retraining to HD-EMG data recorded in separate sessions, muscles, and participants. We recorded HD-EMG signals from the vastus medialis and vastus lateralis muscles while participants performed isometric contractions during two sessions separated by approximately 20 months. We identified motor unit spike trains from HD-EMG signals using a blind source separation (BSS) method, and then used the cumulative spike train (CST) of these motor units and the HD-EMG signals to train and validate the deep CNN. On average, the correlation coefficients between CST from BSS and that from deep CNN were 0.977{+/-}0.007 for leave-one-out across-sessions-and-muscles validation and 0.985{+/-}0.005 for leave-one-out across-participants validation. When trained with more than four datasets, the performance of deep CNN saturated at 0.979{+/-}0.001 for cross validations across muscles, sessions, and participants. Therefore, we can conclude that the deep CNN is generalizable across the afore-mentioned conditions without retraining. We could potentially generate a robust deep CNN to estimate neural drive to muscles for neural-machine interfaces.