Can you see it? Brain-computer interfaces to modulate conscious perception

Poster presented at the conference "Meetings of the Alhambra: The Cognitive Neuroscience of Consciousness" in Granada (Spain), 25-26 April 2025. Abstract: Brain-computer interfaces (BCIs) are a robust and powerful tool that enables the integrationof observed neural activity with direct manipulations of both this activity and behavior. BCIshave been widely used for years in clinical research, particularly with patients suffering frommotor deficits caused by strokes or neurodegenerative diseases. Moreover, BCIs also havepromising applications in basic neuroscience, with recent literature demonstrating theireffectiveness in motor control, mental imagery, and attention.In a previous study we found that relevant information related to conscious perceptionprocesses can be decoded from electroencephalography (EEG) data using machine learning(ML) algorithms. Specifically, our results showed that the presence or absence of a stimuluscould be decoded from the EEG data after target onset. Similar results were also obtained forparticipants’ reported perception. Furthermore, we also observed significantly differentpatterns of activation for present targets depending on whether they were reported byparticipants as seen or as unseen.Based on this perceptual task, we are developing a closed-loop system to detect perceptualfailures (i.e. participants reporting a stimulus as ‘unseen’ when it was presented) to alertparticipants with an attentional signal in order to improve their performance in the task. Thisneurofeedback pipeline includes real-time preprocessing of the recorded EEG signal, trainingof a ML classifier, and deploying the model to continuously analyze registered data. Ourclosed-loop system is able to perform predictions of the performance on the behavioral taskto decide whether to send an alert tone to the participant. Here, we present some preliminaryresults of this neurofeedback system, and lay the groundwork for future steps.