Abstract
Due to protocol complexity and limited recording time, bidirectional brain-computer interfaces (BCIs) that couple neural decoding with artificial somatosensory feedback remain uncommon in clinical settings. We present a proof-of-concept intracranial, bidirectional BCI embedded in virtual reality (VR) and tested in one Epilepsy Monitoring Unit (EMU) participant. The system implements a "gaze & neural trigger" paradigm: gaze selects a virtual object, and a minimalist decoder detects high-gamma bandpower from one intracranial electrode during an overt motor cue to issue a binary grasp command. On contact, electrical stimulation (DES) of somatosensory cortex provides localized neurohaptic feedback (thumb-pressure); a catch trial without DES abolished the percept. Across 12 trials, task accuracy was 83.3% (10 / 12 correct). Mean activation delay was 4.86 \pm 5.47 ~\mathrm{s} ; the first activation required 19 s and fell to \leq 1 ~\mathrm{s} by the seventh trial, remaining \leq 1 ~\mathrm{s} thereafter. During 54 s of non-task behavior, including 25 s of intentional rest and free interaction, the decoder did not trigger-suggesting activation was governed by volitional intent rather than incidental movement. These results demonstrate a clinic-feasible, bidirectional VR-BCI coupling a lightweight single-channel decoder to VR interactions with DES feedback and motivate studies of learning and percept-driven reinforcement.