Artificial intelligence is helping researchers overcome a major hurdle in restoring vision through brain implants by improving how accurately electrical devices stimulate the visual cortex and predict patient perceptions, according to a study published in the journal Neuron.
Improving brain stimulation precision
- Bypassing the Eyes: Visual cortical prostheses send electrical signals directly to the visual cortex at the back of the brain, completely bypassing damaged eyes and optic nerves.
- Smart Neural Adaptation: Instead of using rigid settings, deep learning models analyze real-time brain activity to make artificial vision signals more consistent.
- Targeted Stimulation: The AI approach requires less electrical current while more accurately producing the intended patterns of brain activity.
Testing visual cortical prostheses in Spain
Researchers from the Bionic Vision Lab at the University of California, Santa Barbara (UCSB) traveled to Spain in 2024 to test whether their computer models could enhance an existing experimental brain implant, ScienceDaily reported. Collaborating with teams from ETH Zurich and Miguel Hernández University, the researchers worked at Hospital IMED Elche with a 27-year-old blind man who had lost his sight due to a traumatic brain injury.
The patient had received an experimental implant containing 96 electrodes arranged in an array within his visual cortex. When electrical current activated these electrodes, the participant experienced phosphenes, which are visual sensations that appear as spots, flashes, or shapes of light like stars or fireworks, as detailed by dpaonthenet.net. However, the exact same electrical stimulation does not always trigger an identical response, making predictable vision difficult to achieve.
“Building a model in the abstract is one thing,” said Michael Beyeler, an associate professor of computer science and psychological and brain sciences at UC Santa Barbara.
“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,” Beyeler added.
How deep learning predicts neural responses
To solve the challenge of inconsistent neural firing, the research team trained a deep neural network to predict how the visual cortex would react to various electrical stimulation settings. Insights Integration said that the AI system factored in the brain’s resting activity immediately prior to stimulation to account for fluctuating neural states.
When tested, the AI-selected stimulation patterns reproduced targeted brain activity more accurately than comparison methods while consuming less electrical current. Researchers discovered that neural patterns recorded directly from the brain predicted what the participant actually perceived more accurately than the stimulation parameters themselves.
“Engineers naturally want to treat phosphenes like pixels: stimulate more electrodes, and you should get a more complete image,” Beyeler said.
“But the brain does not work that way. Electrodes interact, neural responses fluctuate, and what we put into the brain is not necessarily what the person perceives. The challenge is to learn more about that transformation,” he explained.
Responding to stimulation
This technology offers a potential pathway for individuals whose visual pathways are damaged but whose visual cortex remains capable of responding to stimulation. Potential candidates include people who have lost their sight due to strokes, neurodegenerative diseases, or traumatic brain injuries.
Beyeler highlighted individuals “who have been able to see for part of their life, but then, because of an inherited eye disease or an accident, have lost their vision,” adding that “for those people, the desire may be very strong to get some vision back.”
“A useful visual prosthesis cannot rely on a fixed recipe,” Beyeler said. “It has to learn how an individual brain responds and adapt the stimulation accordingly. Ultimately, the device should adapt to the person, not the other way around.”
| Trial Parameter | Clinical Detail |
|---|---|
| Participant Profile | 27-year-old man blinded by traumatic brain injury |
| Implant Specifications | 96-electrode array implanted in the visual cortex |
| Primary Technology | Deep neural network predicting cortical responses |
| Collaborating Institutions | UC Santa Barbara, ETH Zurich, Miguel Hernández University |
Seeking medical evaluation
Visual cortical prostheses remain strictly experimental and are not currently available for routine clinical use. Patients with complete degeneration or absence of the visual cortex, or those with active intracranial infections, unstable neurological conditions, or severe bleeding disorders, are generally contraindicated for invasive neurosurgical brain implant procedures. Individuals experiencing sudden vision loss, acute neurological deficits, or head trauma must seek immediate emergency medical evaluation.
Next steps in bionic vision research
The proof-of-concept study demonstrates that personalized, AI-driven adjustments can improve communication between artificial vision hardware and the human brain. While the participant’s experiences were limited to flashes and shapes of light, the findings lay the groundwork for future visual prostheses that continuously adapt to individual neural responses.
The underlying computational research was supported by a five-year, $2 million National Institutes of Health (NIH) Director’s New Innovator Award granted to Beyeler in 2022. The collaborative trial continues to evaluate how adaptive algorithms can make bionic vision systems more reliable.
References
- Moure, P., Granley, J., Grani, F., et al. (2026). Deep learning improves stimulation precision and perceptual prediction in visual cortical prostheses. Neuron.
- The new findings build on earlier computational research supported by Beyeler’s National Institutes of Health (NIH) Director’s New Innovator Award. NIH Director’s New Innovator Award Program.
