A novel explainable machine learning approach for EEG-based brain-computer interface systems

Electroencephalographic (EEG) recordings can be of great help in decoding the open/close hand’s motion preparation. To this end, cortical EEG source signals motor cortex (evaluated 1-s window preceding movement onset) are extracted by solving inverse problem through beamforming. sources epochs used as source-time maps input to a custom deep convolutional neural network (CNN) that is trained perform 2-ways classification tasks: pre-hand close (HC) versus resting state (RE) and open (HO) RE. The developed CNN works well (accuracy rates up \(89.65 \pm 5.29\%\) for HC RE \(90.50 5.35\%\) HO RE), but core present study was explore interpretability provide further insights into activation mechanism during preparation hands’ sub-movements. Specifically, occlusion sensitivity analysis carried out investigate which areas more relevant procedure. Experimental results show recurrent trend spatial across subjects. In particular, central region (close longitudinal fissure) right temporal zone premotor together with primary appear primarily involved. Such findings encourage an in-depth seem play key role