Researchers from UC Santa Barbara, ETH Zurich, and Miguel Hernández University have demonstrated how artificial intelligence could make future visual cortical prostheses—sometimes described as a type of “bionic eye”—more precise and responsive to individual users. The proof-of-concept study used a deep-learning model to design electrical stimulation patterns for electrodes temporarily implanted in the visual cortex of a blind participant. The goal was to better predict how the brain would respond and what the participant would perceive.
The study involved a 27-year-old man who had lost his vision following a traumatic brain injury and had a 96-electrode array temporarily implanted in his visual cortex. Electrical stimulation produced perceptions known as phosphenes, which can appear as spots or shapes of light. Researchers trained an AI model using both stimulation settings and measurements of the participant's brain activity, allowing it to identify stimulation patterns that were more likely to produce targeted neural responses.
When tested on the participant, the AI-designed stimulation patterns reproduced desired patterns of brain activity more accurately while requiring less electrical current than alternative approaches. Importantly, measurements of the brain's actual response were better predictors of what the participant perceived than simply knowing which electrodes had been activated. The researchers also incorporated the participant's resting brain activity, allowing the model to account for changes in neural responses over time.
The findings highlight a major challenge in developing visual prostheses: the brain cannot simply be treated like a screen made of pixels. Neural responses fluctuate, electrodes interact, and identical stimulation does not necessarily produce identical perceptions. AI could therefore help create prostheses that learn how an individual's brain responds and dynamically adjust stimulation. The research is still an early proof of concept, but it represents a step toward more personalized brain-computer interfaces that could eventually help restore limited visual function for some people with severe vision loss.
