Optical Brain Implant Streams ECoG Data at 50 Mb/s Through Skin

A brain-computer interface from Geneva-based Ability Neurotech has recorded human cortical signals through 128 ECoG channels, streaming raw data through the skin at up to 50 megabits per second using an infrared optical link rather than radio-frequency transmission. Last month, the company began an intraoperative study in Germany, capturing neural signals from patients undergoing brain tumor surgery in sessions lasting 20 to 30 minutes.

The implant rests on the brain's surface, using electrocorticography (ECoG) electrode arrays rather than the penetrating electrodes that other experimental BCIs place inside brain tissue. Each channel samples 30,000 times per second. An external headpiece receives the optical data stream, passes it to a separate processor for decoding, and powers the implant wirelessly through induction, eliminating any internal battery that could require replacement.

Ability Neurotech CEO Rotem Kopel said the company evaluated several transmission options before choosing the optical link. The goal was to move raw recordings out of the implant without reduction. The system performs no downsampling or compression before transmission, leaving all decoding and processing to equipment outside the body.

Why raw data matters for decoding

Maitreyee Wairagkar, a project scientist in the Neuroprosthetics Lab at the University of California, Davis, said the battery-free design combined with the optical link "sounds very promising for chronic ECoG recordings."

"Preserving the information available in raw neural data through transmission is also useful for decoding purposes, since precious data is not lost due to transmission limitations," Wairagkar said. "I think this is the right focus for longitudinal BCI use, as it offers flexibility over neural feature extraction, which can be useful for improving decoding performance."

Ability is not alone in pursuing high-bandwidth wireless transmission. Wairagkar noted that other fully implantable BCIs, including both ECoG and penetrating-electrode designs, use similar architectures, with some exceeding 50 Mb/s. "Wireless implantable BCIs are still in early days, and we're seeing multiple new clinical trials being conducted with these devices with varying capabilities in channel counts, signal-to-noise ratio, and data transfer rates, which will determine their performance, functionality, and long-term utility," she said.

What comes next in the trials?

Later stages of the ongoing German study will record brain activity in up to five conscious patients as they perform speech and movement tasks. A separate chronic trial in the Netherlands, planned for late 2026, will test the system over far longer horizons. University Medical Center Utrecht is recruiting participants for a year-long study, part of a project aiming to restore speech for paralyzed individuals, in which people with amyotrophic lateral sclerosis (ALS) will receive the implant, train with it at home, and undergo repeated brain-signal recordings.

Kopel outlined the study's sequence. The company will first evaluate the implantation procedure and the device's safety and performance. The next goal is computer control by participants, followed by speech decoding, translating intended speech from brain activity into words in real time.

The engineering behind the device took roughly 10 years to develop, according to Kopel, followed by about 18 months of bench, durability, and other validation testing after the design was finalized. The implant had to meet tight power and temperature limits; moving large data volumes generates heat, making temperature control a central challenge. Engineers also routed electrode connections into a hermetically sealed case to keep moisture from the electronics, and the optical link must work through varying skin thickness, blood vessels, and hair while tolerating imperfect alignment with the headpiece.

Can the system survive years of daily use?

The sterner test begins after implantation. Software requires updates, brain signals can change over time, and patients may remain dependent on an implant long after the company that built it has changed or vanished.

Wairagkar said sustained performance and decoding accuracy over several years are critical. "Software challenges, like maintaining and calibrating the decoders and updating the user applications, are easier to solve, but it is important that the underlying signal quality obtained from the device is maintained over long periods," she said. Fully implanted wireless systems already avoid one major durability problem: the permanent transcutaneous connection used by some BCIs, whose elimination can reduce infection risk and ease maintenance and everyday use.

Ability has planned for its own potential disappearance. The technology was developed at the nonprofit Wyss Center for Bio and Neuroengineering in Geneva before the company spun out in 2025, and Kopel said that if Ability ceased operations, its intellectual property and responsibilities would return to the center.

"As the field matures, there will need to be structures and policies in place to support the use and deployment of implantable BCIs and to provide appropriate services to patients," Wairagkar said. The late-2026 chronic trial in Utrecht will offer the first extended human test of whether the 128-channel optical design holds its signal quality across a year of daily home use by ALS patients.

By Priya Raman