Real-Time Spatiotemporal Filtering for Artifact-Free EEG during Electrical Neurostimulation

AO_SCPLOWBSTRACTC_SCPLOWCombining electrical neurostimulation with electroencephalography (EEG) for adaptive neurostimulation remains challenging due to the presence of stimulation artifacts in the recorded signal. Interpretation of EEG activity concurrent with stimulation requires real-time filtering of this noisy signal. While traditional frequency domain filters can suppress activity within frequency bands, they fail to differentiate sources in situations when there is a shared frequency characteristic between brain activity and stimulation signal. Here we present a new real-time denoising approach that combines spatial filtering and dynamic filter application. The spatiotemporal filter can suppress stimulation artifacts that share frequency bands with the brain signal in real-time while preserving the full spectral power. The spatial filter is also dynamically updated to account for possible changes in artifact topography. Finally, the filter is robust to changes of stimulation intensity, frequency and duration, enabling reliable denoising performance for inclusion in brain state-based closed-loop stimulation.