AIBRIEF
A UC Santa Barbara led team used deep learning to tune what electrodes in the brain's vision region produce, a peer reviewed step toward devices that bypass the eyes.
A blind participant in Spain sat with an array of electrodes in their visual cortex while a deep-learning model designed, in real time, the electrical pattern that produced a target perception. The result, published this week in Neuron00150-7), is a peer-reviewed proof-of-concept step toward a visual cortical prosthesis, a brain-implant device that bypasses the eyes and optic nerves entirely and stimulates the brain's vision region directly.
The work was led by UC Santa Barbara associate professor Michael Beyeler with co-first authors Pehuén Moure at ETH Zurich, Jacob Granley at UC Santa Barbara, and Fabrizio Grani at Miguel Hernández University, under supervisors Shih-Chii Liu at ETH Zurich and Eduardo Fernández at Miguel Hernández University, according to a UCSB release. The study sits inside a broader feasibility trial underway in Spain.
The model did not restore sight. It gave researchers a per-user tuner for what the brain perceives when electrically stimulated, useful for the population this kind of device is meant to serve: people who lost sight to inherited eye disease, stroke, neurodegenerative disease, or brain injury, and whose visual cortex is still capable of responding. News-Medical's coverage of the paper describes a single-participant demonstration.
The Bionic Vision Lab at UC Santa Barbara frames the result as a step from "theory to something that may one day help people." Clinical translation, including safety, durability, and whether what one participant perceives scales to others, remains open.