Towards a visual prosthesis respecting cortical functional representations

Current vision restoration solutions for patients with a damaged optic pathway rely primarily on the retinotopic position of electrodes to encode visual stimuli. Stimulation protocols that take into account other features (such as orientation) encoded in neurons near stimulation electrodes could allow for more precise control of percepts. However, visually evoked responses cannot be measured in the blind - mapping these features in a straightforward way is impossible. To address this issue, we developed a method for reconstructing the orientation preference map from spontaneous activity on simulated visual cortex data and validated it on macaque recordings. Finally, we demonstrate in blind human patients with a cortical implant that the decoded functional properties influence the evoked percepts.