VR feedback boosts BCI performance by 13% in 3D motor imagery decoding study

A team led by Niall McShane at Ulster University conducted the first systematic investigation of embodied virtual reality (VR) feedback during real-time 3D virtual limb control driven solely by motor imagery, with ten participants across ten longitudinal sessions. They used a CNN-LSTM decoder to decode imagined movement trajectories in three dimensions, comparing VR feedback to standard screen feedback. The VR condition achieved significantly higher within-session imagined movement correlations (r=0.762 vs r=0.672, p≤0.002, Cohen's d=1.42–2.05), representing an 8.9–13% improvement across all movement axes and evaluation strategies (fixed decoders, periodic retraining, and within-session reconstruction).

Neurophysiologically, VR feedback elicited stronger sensorimotor-parietal desynchronisation, enhanced motor-frontal functional connectivity, and pervasive anterior insula engagement across all frequency bands, paralleling patterns seen during actual movement. These effects persisted even without decoder retraining, indicating that embodied VR feedback inherently produces more decodable and generalisable neural representations. The findings establish embodied spatial feedback as a key design principle for next-generation continuous brain-computer interfaces targeting intuitive motor control, neurorehabilitation, and potentially prosthetic or exoskeleton control. The paper has been submitted to Nature Biomedical Engineering and data will be made available via Zenodo.