EEG-guided VR training fails to boost prosthetic vision performance in study

Visual prostheses require users to actively interpret sparse, distorted artificial percepts, making effective training critical. To address this, researchers from the University of California, Santa Barbara and McMaster University developed a closed-loop EEG feedback platform that adjusts training in real-time based on users' cognitive engagement. The system integrates a dry-electrode EEG headset with a VR head-mounted display and was tested on 22 sighted participants searching a virtual desk scene rendered through a low-resolution phosphene simulation.

During the object-localization task, one group received post-trial visual feedback based on a commonly used EEG engagement index (the ratio of β to α+θ brainwave activity), while a control group received visually matched but non-contingent sham feedback. Surprisingly, both groups showed comparable within-session improvements in localization performance, likely due to practice, increased familiarity with the simulated percepts, or refined search strategies. The EEG-contingent feedback did not produce reliable group-level benefits in accuracy, completion time, workload, or modulation of the targeted index.

Exploratory analyses revealed substantial individual variability, but no feedback-specific relationship between the engagement index and behavioral performance. The study demonstrates the feasibility of integrating EEG-contingent feedback with immersive simulated prosthetic vision, but identifies important limitations: the EEG measure may be too noisy or slow for feedback in a task that changes rapidly as participants learn, and the single-session training may have been too short to allow feedback effects to emerge.