Predicting Somatosensory Awareness from Intracranially Informed Scalp EEG

Abstract Behavioural reports are the standard means of determining whether a stimulus is consciously perceived, but assessing perception when patients are unable to provide such reports remains challenging. Here, using a machine-learning approach trained on simultaneous high-density scalp and intracranial EEG recordings during median nerve stimulation, we derived a candidate non-invasive marker of somatosensory awareness by predicting from scalp EEG the intracranial tonic operculo-insular response previously associated with conscious somatosensory perception. In an independent multicentre perceptual-threshold dataset of healthy participants, the scalp estimate of tonic activity was larger for perceived than unperceived stimuli and consistently distinguished the two conditions at the individual level. This participant-level consistency was not observed for the Somatosensory Awareness Negativity, the best-established scalp EEG marker of somatosensory awareness. Moreover, a decision threshold derived from the perceptual dataset generalized without recalibration to a third, independent dataset, separating wakefulness from deep NREM sleep across all participants despite the absence of concurrent perceptual reports. Thus, our approach identifies an intracranially informed, non-invasive EEG marker associated with conscious somatosensory processing that generalizes across individuals, datasets and brain states, supporting its future evaluation in clinical settings where behavioural reports are unavailable.