A new error-monitoring brain-computer interface based on reinforcement learning for people with autism spectrum...

Area of Science:

• Neuroscience
• Computer Science
• Psychology

Background:

• Brain-computer interfaces (BCIs) offer tailored feedback for cognitive training in neurodevelopmental disorders.
• Autism spectrum disorder (ASD) often involves altered error monitoring processes, making traditional interventions challenging.

Purpose of the Study:

• To propose and validate a gamified BCI using non-volitional neurofeedback for cognitive training in ASD.
• To assess the feasibility of a BCI system where users implicitly train an agent by detecting errors.

Main Methods:

• Developed a gamified BCI with an emotional facial expression paradigm controlled by a reinforcement learning (RL) agent.
• Participants observed and judged the agent's actions, generating error-related potentials (ErrPs) upon incorrect actions.
• Tested the BCI with neurotypical participants and one participant with ASD to evaluate online ErrP detection and agent learning.

Main Results:

• Achieved online balanced accuracy for ErrP detection of 81.6% and 77.1% in two game modes.
• All participants successfully trained the RL agent to an optimal strategy in at least one session.
• Demonstrated the feasibility of the BCI methodology for clinical application in ASD.

Conclusions:

• The developed BCI system shows promise as a neurorehabilitation tool for ASD.
• Successful ErrP detection and agent learning validate the proposed approach for further clinical investigation.