Enhancing Motor Imagery Performance with Real-time Augmented Reality Feedback

Abstract

Immersive head-mounted augmented reality (AR) provides a novel form of visual feedback in motor imagery-based brain-computer interfaces (MI-BCIs), enhancing user immersion. To date, a closed-loop, real-time, electroencephalography-driven MI-BCI with AR visuals has not been demonstrated. In this study, a real-time AR MI-BCI was used, in which hand-tracked virtual hands were rendered over participants’ own hands and animated according to classifier output. This system was compared to conventional on-screen abstract-bar feedback in a within-subject crossover experiment with 10 healthy participants performing left-right hand-grasping motor imagery. Across three real-time feedback blocks, average classification accuracy was nearly identical between conditions (62.85% AR vs. 62.87% control), but only the AR condition showed a statistically significant improvement from the first to later blocks (p<0.05), indicating better adaptation to feedback with repeated training. AR sessions also produced significantly greater event-related desynchronization in later blocks, consistent with increased sensorimotor engagement, while simulator-sickness scores remained minimal. These findings establish the feasibility of real-time AR feedback for MI-BCI and suggest potential benefits for extended, multi-session neurorehabilitation training.