Insular error network enables self-correcting intracranial brain-computer interface

Abstract

Error recognition is fundamental to adaptive behavior, enabling rapid compensatory action when outcomes deviate from expectations. Central to this function are neural circuits for performance monitoring, encoding cognitive signals that could support more reliable neural interfaces. Here, we recorded intracranial electroencephalography (iEEG) in epilepsy patients to enable a motor brain-computer interface (BCI) while sampling error-related activity across a distributed network. Our work reveals high-frequency population dynamics emerging in the anterior insula and propagating to the prefrontal cortex as the interface fails to follow the users intention. We identify spatially organized insular responses to error processing and movement feedback, highlighting it as a heterogeneous hub linking action and outcome. Real-time integration of error responses enables a self-correcting neural interface that enhances usability by reducing the need for manual user intervention. Together, our work demonstrates a human intracranial BCI harnessing insular brain activity, integrating cognitive processes directly into device control.