New tool brings us one step closer to visual implants

17 March 2025

Researchers at the Netherlands Institute for Neuroscience have collaborated with several universities to develop a new open-source tool that can be used to optimise the placement of visual brain implants on a large scale.

Approximately 40 million people suffer from blindness worldwide, a number that is expected to increase over the coming years. Pieter Roelfsema’s group is working on electrodes that can be implanted directly onto the brain’s visual cortex, which could eventually restore a rudimentary form of vision.

Antonio Lozano, co-author and researcher in the Roelfsema group, has built a tool that can optimise the placement of such an implant. “Everyone’s brain is different, so not every array or chip-design will be suitable for every person”, he explains. “The location of an implant must therefore be adapted for each individual case.”

Globally, various research groups are developing brain implants. “We’re getting closer to tangible results, but we are not the only ones. Companies such as Neuralink are progressing very fast well. Once these implants reach clinical trials, there will be so many different models and channels that surgical availability could become a major bottleneck”, Lozano explains.

A personalised approach

Choosing where to implant the prosthesis depends on various factors, but Lozano’s tool incorporates several important ones. The first is the total area that needs to be covered (i.e., how large is the implant?). Another important factor is the yield: how can the implant be placed in such a way that it reaches as many nerve cell bodies as possible? The third factor depends on the way an implant sends its signal to the brain. Some implant designers may decide to mimic the human eye by creating implants with a built-in focus point. In this case the signal from the implant is more concentrated in the centre and less near the periphery. Other companies may create an implant that distributes the signal evenly so that someone has a better idea of the entire environment, but no clear focus point.

The tool also incorporates a safety constraint. “This is probably the most critical part”, Lozano emphases while pointing to a large group of veins covering the brain’s visual cortex. “Damaging any veins must be avoided at all costs. So our algorithm searches for the sweet spot that optimises the output while minimising the safety risks”, he explains.

A potential safe location for a visual implant (in blue) while the veins (pink) covering the visual cortex remain intact.

An effective algorithm for artificial vision

To identify the optimal location, Lozano’s tool incorporates a dataset from Noah Benson, a senior data scientist in New York. Using the dataset, the tool can predict what someone will see when the implant is placed on a specific location. If this process is repeated over several locations, the tool can identify the optimal implant position for the person’s specific visual needs.

The steps taken by the tool can be compared to those taken by a surgeon, who would identify a person’s needs and predict the optimal location through a strenuous trial-and-error process. “Our tool does not require much computing power, but it is very efficient”, Lozano adds. “A surgeon could use this dataset from New York, reach real predictions, and optimise the implant location for hundreds of people.”

Since the data is based on participants with intact vision, it remains unclear what a blind person will be able to see with such an implant. For research purposes, Lozano’s group has built a VR-model in which they incorporate all available information about the stimulation process. “We still can’t know with absolute certainty how the brain responds to external stimulation, but this model is our best possible prediction”.

Large-scale and effective implementation

A crucial aspect of Lozano’s tool is that it is freely available. “This is the key”, he adds. “We want other labs and organisations to be able to use our tool. That way, anyone who has their brain scanned can access it and easily identify where the implant can be placed for safe and optimal results.”

Even though the tool is already usable and openly accessible, Lozano will continue to develop his tool further. This update could, for example, take additional implants or ‘flexible threads’ into consideration. “Neuroscience works at different speeds. Since my work only requires a computer, I can work ahead. Plus, it’s a very fun process”, he chuckles.

Still, Lozano hopes the current tool will prove to be effective once implants are ready for clinical trials. “We want to make this technology as useful and as safe as possible for many people. That’s why we’re emphasising its large scale. And it’s not just the big numbers. With our tool, neurosurgeons can optimise brain implants design effectively. I definitely think this will be very useful”.

Click here to view the infographic