👁️ Neuroscience • Neuroprosthetics • AI • Visual Prosthesis • Blindness

AI Helps More Precisely Control Visual Prosthesis in a Blind Person

📋 UMH • ETH Zurich • UCSB • Cortical Visual Prosthesis • Phosphenes • BioRxiv • 6 min read

✅ Reviewed by Dr. Robert Mitchell, PharmD — Clinical Pharmacist & Medical Writer

👁️ Researchers have used machine learning to improve the control of a cortical visual prosthesis, generating more stable phosphenes with less electrical current in a blind participant.

👁️ The Breakthrough — AI Enhances Visual Prosthesis Control

In Spain, scientists from the Miguel Hernández University (UMH), in collaboration with colleagues from ETH Zurich and the University of California, Santa Barbara (UCSB), have developed an algorithm for a cortical visual prosthesis. In a blind participant, the AI was able to reproduce targeted activity patterns in the visual cortex more accurately than standard calibration, while using less electrical current.

The study, published as a preprint on bioRxiv, demonstrates the potential of machine learning to improve the stability and precision of visual prostheses for blind individuals.

🧠

Cortical Implant

96-microelectrode array in visual cortex

⚡

Less Current

AI worked with lower stimulation intensity

💡

Stable Phosphenes

More consistent visual perceptions

💡 Key Finding: The AI algorithm was able to predict and reproduce targeted neural activity patterns with higher precision than standard calibration methods — while using less stimulation current.

🔬 How the Cortical Visual Prosthesis Works

Unlike retinal implants, which are placed in the eye, this type of prosthesis is implanted directly into the visual cortex of the brain. An array of 96 microelectrodes delivers electrical pulses and simultaneously records neuronal responses in real time.

🧬

96-Electrode Array

Microelectrodes are implanted in the visual cortex — they both stimulate and record neural activity.

🎯

Targeted Stimulation

Electrical impulses create phosphenes — flashes or spots of light that the brain interprets as visual information.

🤖

Machine Learning

The model learns from the recorded neural responses to optimize stimulation patterns for each electrode.

📌 Key Insight: The prosthesis is not an artificial eye — it bypasses the eye entirely and directly stimulates the brain’s visual cortex, making it a promising option for people with damage to the eyes or optic nerve.

💡 What Are Phosphenes?

Phosphenes are visual perceptions — flashes, spots, or patterns of light — that are not caused by light entering the eye. They are artificially generated by electrical stimulation of the visual cortex. For people with a cortical visual prosthesis, phosphenes are the “pixels” of artificial vision.

💡

Stable Phosphenes

AI helped generate phosphenes that were more stable and consistent than with standard calibration.

⚡

Less Current

The AI approach achieved better results with lower electrical current — potentially safer for long-term use.

🎯

Precise Patterns

The AI learned to determine which combinations of pulses produce the desired neural response across multiple electrodes.

⚠️ Limitations — What This Study Does Not Show

While promising, the researchers emphasize several important limitations:

👤

Single Participant

The experiment was conducted with only one blind participant — results may not generalize to others.

📄

Preprint, Not Peer-Reviewed

The work has been published as a preprint on bioRxiv and has not yet undergone formal peer review.

👁️

Real-World Tasks

The study did not test whether the improved phosphene control helps the participant read, navigate, or recognize objects.

⚠️ Critical Note: This is a proof-of-concept study with a single participant. The next step is to conduct research with a larger number of people to validate these findings and assess real-world functional benefits.

🔬 Next Steps — Toward Adaptive Visual Prostheses

This approach could become the foundation for adaptive visual prostheses that automatically adjust to the user’s brain activity. However, next steps include:

👥

Larger Studies

Testing the approach in more participants to confirm efficacy and safety.

👁️

Real-World Tasks

Assessing whether improved phosphene control translates to reading, navigation, and object recognition.

🤖

Adaptive Algorithms

Developing AI models that continuously adapt to changes in brain activity over time.

💡 Key Insight: “This is a proof of concept,” the researchers emphasize. “But it suggests that machine learning could be a powerful tool for improving the performance of neural prostheses.”

❓ Frequently Asked Questions

What is a cortical visual prosthesis?

A cortical visual prosthesis is a device implanted directly into the visual cortex of the brain. It bypasses the eyes and optic nerve, delivering electrical pulses that create phosphenes — visual perceptions that can help blind individuals perceive their environment.

How did AI improve the visual prosthesis?

The AI learned to predict and reproduce targeted neural activity patterns more accurately than standard calibration methods. It achieved this while using less electrical current, potentially making the device safer for long-term use.

What are phosphenes?

Phosphenes are flashes, spots, or patterns of light that are perceived without light entering the eye. In visual prostheses, they are artificially created by electrical stimulation of the visual cortex.

Does this mean blind people can see again?

Not yet. This is a proof-of-concept study with one participant. The AI improved the stability and precision of phosphenes, but the study did not test whether this helps with reading, navigation, or object recognition. Much more research is needed.

Is this treatment available to patients now?

No. This is an experimental approach that has only been tested in one participant. The research is still in early stages and has not yet been peer-reviewed. It is not currently available to patients.

📚 References & Resources

⚠️ Medical Disclaimer

This content is for informational and educational purposes only. The research findings discussed are preliminary and based on a preprint study — they have not yet been peer-reviewed and have only been tested in one participant. The technology is not currently available as a clinical treatment. This information does not constitute medical advice. Always consult a qualified healthcare professional for personalized health guidance.

Dr. Tina Sugandh

All Author Posts

All Departments

Recently Viewed