Researchers from three institutions, including the University of California, Santa Barbara, announced on Aug. 7 that artificial intelligence may improve the precision and predictability of future visual prostheses such as bionic eyes. Michael Beyeler, associate professor of computer science at UCSB, and colleagues used a deep-learning model to design patterns of electrical stimulation for electrodes temporarily implanted in the visual cortex—the part of the brain that processes visual information—of a blind participant. The model allowed researchers to better control how neurons responded to stimuli and helped predict what the participant perceived.
The study, published online in Neuron, is described as a step toward developing visual cortical prostheses that can communicate more effectively with the brain. "Building a model in the abstract is one thing," said Beyeler. "Seeing it shape an experiment with a person is something else entirely. That ability to go from theory to something that may one day help people is what drives much of the work in our lab." The project was led by co-first authors Pehuén Moure (ETH Zurich), Jacob Granley (UCSB), and Fabrizio Grani (Miguel Hernández University). Supervisors included Beyeler, Shih-Chii Liu (ETH Zurich), and Eduardo Fernández (MHU). The research forms part of an ongoing feasibility trial in Spain.
Visual prosthesis development has been attempted for decades with limited success. Some devices target the retina while others stimulate later stages within the visual system. Cortical prostheses bypass both eyes and optic nerves by delivering electrical stimulation directly to the brain's visual cortex. This approach could benefit individuals whose blindness results from strokes, neurodegenerative diseases or injuries but who still have responsive visual cortices.
The current work builds on computational research funded by Beyeler's 2022 National Institutes of Health Director's New Innovator Award—a five-year grant supporting efforts to make these devices more predictable and effective. In 2024, members of Beyeler’s Bionic Vision Lab traveled to Hospital IMED Elche in Spain, where they worked with a 27-year-old man who had lost his vision due to traumatic brain injury and received an implant composed of a 96-channel electrode array.
Using their AI model trained on neural activity data before each test session, researchers identified stimulation patterns likely to produce desired neural responses. When tested on the participant, these patterns reproduced targeted brain activity more accurately using less electrical current than other approaches; moreover, recorded neural activity predicted perception better than just knowing which electrodes were activated.
Beyeler said, "A useful visual prosthesis cannot rely on a fixed recipe... Ultimately, the device should adapt to the person, not the other way around."
