Neural mechanisms of training in Brain-Computer Interface: A Biophysical modeling approach

Brain-computer interface (BCI) is a system that translates neural activity into commands, allowing direct communication between the brain and external devices. Despite its clinical application, BCI systems fail to robustly capture subjects’ intent due to a limited understanding of the neural mechanisms underlying BCI control. To address this issue, we introduce a biophysical modeling approach that leverages a linear neural mass model to investigate the associated neural mechanisms of motor imagery-based BCI experiments. We tailor this model to simulate both motor imagery task and resting state. We apply this approach to a cohort of 19 healthy subjects trained along four sessions where magnetoencephralography (MEG) and electroencephalography (EEG) signals were simultaneously recorded. The neural synaptic gain and time scale of the modeled excitatory and inhibitory neural mass populations capture changes in neural activity across conditions and sessions. Those changes appear in important areas of the sensorimotor cortex, relevant for motor imagery tasks. We observed these effects in both EEG and MEG modalities. These findings provide insights into the underlying neural mechanisms in a motor imagery task in BCI, paving the way to tailored BCI training protocols.