MEGNet: A MEG-Based Deep Learning Model for Cognitive and Motor Imagery Classification

Decoding complex patterns associated with task-specific activities embedded within magnetoencephalography (MEG) signals is pivotal for understanding brain functions and developing applications such as brain-computer interfacing. It is widely recognized that machine learning algorithms rely on feature extraction before undertaking decoding tasks. In this work, we introduce MEGNet, aiming to enhance the single-trial decoding framework of a compact deep neural network inspired by EEGNet, a model widely utilized in electroencephalography (EEG) studies. MEGNet accepts raw MEG signals, evoked responses and frequency spectrum as input. For validation, the MEG dataset containing motor and cognitive imagery tasks was used for classification. We performed pair-wise decoding of cognitive and motor tasks. Classification accuracy was evaluated using metric scores and benchmarked against ShallowConvNet and DeepConvNet. Our findings demonstrate that MEGNet can successfully decode between cognitive and mental imagery tasks. This MEGNet model surpasses existing feature extraction techniques, exhibiting consistent and stable mean accuracy of 64.76%±3% across tasks and subjects.