Artificial intelligence models can successfully optimize electrical stimulation in visual cortical prostheses, improving the accuracy and efficiency of artificial vision, according to a study published in Neuron by researchers from UC Santa Barbara, ETH Zurich, and Miguel Hernández University.
Visual cortical prostheses, commonly known as bionic eyes, bypass the eyes and optic nerves completely. Instead, these devices deliver electrical stimulation directly to the visual cortex located at the back of the brain. This technology provides a potential option for individuals who have lost their sight due to stroke, traumatic brain injury, or neurodegenerative disease, provided they retain a functional visual cortex.
Overcoming Mechanical Challenges in Bionic Eyes
Traditional visual prostheses face significant mechanical hurdles that limit their effectiveness. The human brain does not process electrode signals as simple pixels on a computer screen. Instead, adjacent electrodes interact with one another, and neural responses naturally fluctuate over time.
Standard electrical stimulation settings also suffer from a perceptual disconnect. These rigid settings fail to reliably predict what a user actually perceives, such as phosphenes, which are spots of light generated by the stimulation. According to the research team, static protocols cannot maintain reliable artificial vision because neural responses change on a daily basis.
Did you know?
Instead of treating the brain like a standard pixel grid, modern deep-learning models account for the fact that adjacent electrodes interact and neural responses shift throughout the day.
Testing Deep-Learning Models in Human Trials
To address these hurdles, researchers tested a deep-learning model on a 27-year-old blind participant in Spain who was implanted with a 96-channel cortical electrode array. Rather than relying on fixed electrical parameters, the team trained a deep neural network on actual brain activity responses. This model incorporated the participant’s resting brain state immediately prior to stimulation.
The AI-designed stimulation patterns reproduced target brain activity much more accurately while demanding significantly lower electrical current. Furthermore, recorded neural responses served as a far stronger predictor of perceived phosphene features—including shape, size, brightness, and colour—than raw electrode settings alone.
Adaptive Systems for Long-Term Usability
By combining resting-state measurements with closed-loop neural feedback, the deep-learning framework allows the prosthesis to adapt to shifting brain states in real time. This approach shifts the technology away from rigid, one-size-fits-all protocols.
“A useful visual prosthesis cannot rely on a fixed recipe,” said Michael Beyeler. “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.”
Frequently Asked Questions
How do visual cortical prostheses work?
Why is AI necessary for bionic eyes?
Because the brain does not process signals like simple pixels and neural responses fluctuate daily, AI models help optimize electrical stimulation for better accuracy and lower energy use.
What are phosphenes?
Phosphenes are the spots of light perceived by users when electrical stimulation activates neurons in the visual cortex.
Want to stay updated on the latest breakthroughs in neurotechnology and artificial intelligence? Subscribe to our newsletter or leave a comment below to join the discussion.
Keep reading