Enhanced Spatial Division Multiple Access BCI Performance via Incorporating MEG With EEG

Abstract

Spatial division multiple access (SDMA) is a way of encoding BCI systems based on spatial distribution of brain signal characteristics. However, SDMA-BCI based on EEG had poor system performance limited by spatial resolution. MEG-EEG fusion modality analysis can help solve this problem. According retina-cortical relationship, this study used stimulus out of the central visual field and tiny fixation points to construct a 16-command SDMA coded MEG-EEG fusion modality BCI system. We achieved this by synchronously acquiring MEG and EEG signals from 10 subjects. We compared the spatiotemporal features between MEG and EEG by analyzing signals in the occipital region. We fused MEG and EEG modalities without any signal processing and used the multi-class discriminative canonical pattern matching (Multi-DCPM) algorithm to evaluate and compare the system performance of EEG, MEG, and MEG-EEG fusion modalities. The result showed that MEG and EEG had obvious differences in spatial distribution characteristics. MEG improves offline classification accuracy of the 16 fixation points by 27.81% over EEG at 4s data length. Specially, the MEG-EEG fusion modality achieves an impressive average offline accuracy of 91.71%, which was a significant improvement over MEG (p<0.01, ANOVA). The MEG-EEG fusion modality achieved average information transfer rate (ITR) of 60.74 bits/min with a data length of 1 s, which was a 14% improvement over MEG. The MEG-EEG fusion modality significantly enhanced the spatial features and performance of SDMA-encoded BCIs. These results highlight the potential and feasibility of MEG-EEG fusion modality BCI, and provide theoretical insights and practical value for promoting the further development and application of SDMA in BCI.