Electroencephalography-based a Motor Hotspot Identification Approach Using Deep-Learning

Neurorehabilitation based on transcranial electrical stimulation (tES) has been introduced to improve the motor rehabilitation for patients with neurological disorders. To define an optimal tES site, transcranial magnetic stimulation (TMS) is generally used. However, although TMS is an optimal tool to identify an individual motor hotspot for tES, it requires a cumbersome procedure involving the empirical judgment of an expert. In this study, we proposed a convolutional neural network (CNN)-based motor hotspot identification approach using electroencephalography (EEG). EEG were recorded from twenty subjects while they repeatedly performed a simple finger-tapping task using the right index finger. The power spectral densities (PSDs) were extracted as features. The PSD features of each channel and the 3D coordinates of the motor hotspot identified by TMS-induced MEP were used as input features to train the CNN model. A minimum error distance between the motor hotspots identified by TMS and EEG was 0.04 cm, demonstrating the feasibility of our proposed EEG-based motor hotspot detection method.