Haslacher, D. | bioRxiv Science

bioRxiv2020

Stimulation Artifact Source Separation (SASS) for assessing electric brain oscillations during transcranial alternating current stimulation (tACS)

Brain oscillations, e.g. measured by electro- or magnetoencephalography (EEG/MEG), are causally linked to brain functions that are fundamental for perception, cognition and learning. Recent advances in neurotechnology provide means to non-invasively target these oscillations using frequency-tuned amplitude-modulated transcranial alternating current stimulation (AM-tACS). However, online adaptation of stimulation parameters to ongoing brain oscillations remains an unsolved problem due to stimulation artifacts that impede such adaptation, particularly at the target frequency. Here, we introduce a real-time compatible artifact rejection algorithm (Stimulation Artifact Source Separation, SASS) that overcomes this limitation. SASS is a spatial filter (linear projection) removing EEG signal components that are maximally different in the presence versus absence of stimulation. This enables the reliable removal of stimulation-specific signal components, while leaving physiological signal components unaffected. For validation of SASS, we evoked brain activity with known phase and amplitude using 10 Hz visual flickers across 7 healthy human volunteers. 64-channel EEG was recorded during and in absence of 10 Hz AM-tACS targeting the visual cortex. Phase differences between AM-tACS and the visual stimuli were randomized, so that steady-state visually evoked potentials (SSVEPs) were phase-locked to the visual stimuli but not to the AM-tACS signal. For validation, distributions of single-trial amplitude and phase of EEG signals recorded during and in absence of AM-tACS were compared for each participant. When no artifact rejection method was applied, AM-tACS stimulation artifacts impeded assessment of single-trial SSVEP amplitude and phase. Using SASS, amplitude and phase of single trials recorded during and in absence of AM-tACS were comparable. These results indicate that SASS can be used to establish adaptive (closed-loop) AM-tACS, a potentially powerful tool to target various brain functions, and to investigate how AM-tACS interacts with electric brain oscillations. Highlights- Stimulation Artifact Source Separation (SASS), a real-time compatible signal decomposition algorithm for separating electric brain activity and stimulation signal artifacts related to amplitude-modulated transcranial alternating current stimulation (AM-tACS), is introduced - Employing SASS, phase and amplitude of single-trial steady state visual evoked potentials (SSVEPs) were reliably recovered from electroencephalography (EEG) recordings at the frequency targeted with AM-tACS - SASS enables assessment of single-trial oscillatory brain activity at the target frequency during stimulation and paves the way for online adaptation of stimulation parameters to ongoing brain oscillations

neuroscience↗

bioRxiv2020

Minimizing biosignal recording sites for hybrid noninvasive brain/neural robot control

Noninvasive brain/neural controlled robots are promising tools to improve autonomy and quality of life in severe paralysis, but require biosignal recordings, such as electroencephalography (EEG) and electrooculography (EOG), from various sites distributed over the users head. This limits the applicability and practicality of noninvasive brain/neural robot control on an everyday basis. It would thus be very desirable to minimize the number of necessary recording sites paving the way for miniaturized, headset-like EEG/EOG systems that users with hemiplegia can mount by themselves. Here, we introduce a novel EEG/EOG brain/neural robot control strategy using only scalp electrodes placed near cortical sensorimotor areas. The strategy was tested across 16 healthy volunteers engaging in an EEG/EOG brain/neural control task. Classification accuracies were compared using scalp electrodes only vs. the conventional electrode placements across the scalp and face. To evaluate whether cranial muscle artifacts impede classification accuracy, participants were asked to chew during the task. We found that brain/neural classification accuracy was comparable and that chewing did not impact classification accuracies when using scalp electrodes only. Our results suggest that the proposed new strategy allows for reliable EEG/EOG-based brain/neural robot control, a critical prerequisite to broaden the use of noninvasive brain/neural assistive and rehabilitative technologies.

neuroscience↗