ABSTRACT
Magnetoencephalography (MEG) is a non-invasive functional imaging technique for pre-surgical mapping. However, movement-related MEG functional mapping of primary motor cortex (M1) has been challenging in presurgical patients with brain lesions and sensorimotor dysfunction due to the large numbers of trails needed to obtain adequate signal to noise. Moreover, it is not fully understood how effective the brain communication is with the muscles at frequencies above the movement frequency and its harmonics. We developed a novel Electromyography (EMG)-projected MEG source imaging technique for localizing M1 during ~1 minute recordings of left and right self-paced finger movements (~1 Hz). High-resolution MEG source images were obtained by projecting M1 activity towards the skin EMG signal without trial averaging. We studied delta (1-4 Hz), theta (4-7 Hz), alpha (8-12 Hz), beta (15-30 Hz), and gamma (30-90 Hz) bands in 13 healthy participants (26 datasets) and two presurgical patients with sensorimotor dysfunction. In healthy participants, EMG-projected MEG accurately localized M1 with high accuracy in delta (100.0%), theta (100.0%), and beta (76.9%) bands, but not alpha (34.6%) and gamma (0.0%) bands. Except for delta, all other frequency bands were above the movement frequency and its harmonics. In both presurgical patients, M1 activity in the affected hemisphere was also accurately localized, despite highly irregular EMG movement patterns in one patient. Altogether, our EMG-projected MEG imaging approach is highly accurate and feasible for M1 mapping in presurgical patients. The results also provide insight into movement related brain-muscle coupling above the movement frequency and its harmonics.
ABSTRACT
Advancements in deep learning algorithms over the past decade have led to extensive developments in brain-computer interfaces (BCI). A promising imaging modality for BCI is magnetoencephalography (MEG), which is a non-invasive functional imaging technique. The present study developed a MEG sensor-based BCI neural network to decode Rock-Paper-scissors gestures (MEG-RPSnet). Unique preprocessing pipelines in tandem with convolutional neural network deep-learning models accurately classified gestures. On a single-trial basis, we found an average of 85.56% classification accuracy in 12 subjects. Our MEG-RPSnet model outperformed two state-of-the-art neural network architectures for electroencephalogram-based BCI as well as a traditional machine learning method, and demonstrated equivalent and/or better performance than machine learning methods that have employed invasive, electrocorticography-based BCI using the same task. In addition, MEG-RPSnet classification performance using an intra-subject approach outperformed a model that used a cross-subject approach. Remarkably, we also found that when using only central-parietal-occipital regional sensors or occipitotemporal regional sensors, the deep learning model achieved classification performances that were similar to the whole-brain sensor model. The MEG-RSPnet model also distinguished neuronal features of individual hand gestures with very good accuracy. Altogether, these results show that noninvasive MEG-based BCI applications hold promise for future BCI developments in hand-gesture decoding.
