A high-resolution brain interface records movement signals from the brain’s surface, enabling real-time control performance similar to invasive implants without entering brain tissue.
(Nanowerk Spotlight) When the connection between brain and body is disrupted by injury or disease, the signals that once guided movement remain active but trapped. The person may still imagine reaching for a glass or lifting a foot, but the pathway to act on that intention is gone. Restoring that connection without relying on muscles has become one of the most technically and ethically complex problems in neuroscience. At the heart of it is a question of access: how to record neural activity with enough resolution to translate it into useful commands, and how to do that safely, reliably, and in real time.
The most precise brain computer interfaces today use electrodes implanted directly into brain tissue. These systems can record from individual neurons, offering sharp detail and low delay. But they also come with tradeoffs. The body treats them as foreign, scar tissue builds up, and long term reliability becomes uncertain.
At the opposite end are noninvasive systems that use scalp electrodes to measure brain waves, but these signals are weak and slow, blurred by the layers of bone and tissue they pass through. The middle ground, recording directly from the brain’s surface without entering it, has remained promising but underdeveloped.
Electrocorticography, which captures brain activity from electrodes placed on the cortex, offers a path toward this middle ground. It avoids penetrating the brain yet records signals that are faster and more localized than scalp based methods. But clinical systems based on this approach have been limited by electrode size, spacing, and hardware constraints that restrict their resolution and performance. Recording from smaller areas of the brain with more precision, without increasing surgical risks, has remained difficult.
Advances in flexible electronics and microfabrication have made it possible to build high density arrays that conform to the brain’s surface with minimal intrusion. These devices, known as micro electrocorticography or μECoG arrays, can sample brain activity at much finer spatial scales. What has remained unclear is whether they can maintain stable, high quality recordings over time and whether that added precision translates into better control of external devices.
A new study in Advanced Science (“Chronically Stable, High-Resolution Micro-Electrocorticographic Brain-Computer Interfaces for Real-Time Motor Decoding”) addresses these questions directly. The researchers developed a high density μECoG brain computer interface designed to decode motor signals with high accuracy while minimizing invasiveness. They tested the system over more than six months in a large animal model, and then in human trials, where participants used the interface to control computer based tasks with both real and imagined movement.
The results point to a system that offers the signal quality of deeply implanted electrodes while avoiding the biological and surgical risks those systems often carry.
Design and long-term validation of the high-resolution μECoG brain-computer interface system. (a) Schematic of the system workflow showing real-time motor and motor imagery decoding based on signals recorded from a flexible surface electrode array. (b) Photograph of the integrated implant, including the μECoG electrode array, signal processing unit, and protective titanium enclosure. (c) Electrode impedance measurements over a 203-day in vivo experiment in a canine model show minimal signal degradation, with less than 6 percent loss of electrode functionality. (d) Signal-to-noise ratio (SNR) remained consistently above 20 decibels across sessions, indicating stable high-quality recordings. (e) Decoding accuracy, measured by correlation between predicted and actual limb trajectories, remained stable across three movement dimensions. (f) Postmortem histological analysis showed no significant neuronal loss or inflammation in tissue under the implanted array. (g) Quantification of immune markers confirmed that microglial and astrocyte responses in the implanted region were comparable to control tissue. These results demonstrate the system’s chronic stability, biological safety, and suitability for real-time motor decoding over extended periods. (Image: Reprinted from DOI:10.1002/advs.202506663, CC BY) (click on image to enlarge)
The μECoG array developed for this study uses 64 electrodes per square centimeter, embedded in a flexible polyimide structure. Each electrode is 850 micrometers in diameter with 1250 micrometers of spacing between them. These arrays were fabricated using microelectromechanical systems techniques and designed to match the brain’s surface with high conformity. A signal processing unit, connected through a flexible printed circuit and enclosed in a titanium case, enables long term use while protecting the electronics from moisture and mechanical stress.
The researchers validated the system in a Labrador dog, recording brain signals over a 203 day period during natural movement. Throughout the experiment, the device maintained strong signal quality, with signal to noise ratios above 20 decibels and a minimal drop in functioning electrodes. The predicted motion trajectories matched actual joint movements with a mean decoding accuracy of 0.84. The system consistently captured motion related brain activity across three dimensions and retained this performance over time. No evidence of inflammation or neuron loss was found in the implanted brain tissue after the experiment, confirming the biological safety of the implant.
The study also explored how electrode density influences decoding quality. By comparing different subsets of the array, the researchers showed that increasing density improved accuracy and reduced variability in predictions. The optimal density was found at 64 electrodes per square centimeter, with no further gain beyond that point. Conversely, reducing the physical coverage of the array, while keeping density constant, caused decoding performance to decline. These results suggest that high density arrays can reduce the need for extensive brain coverage, allowing for smaller surgical openings while preserving control performance.
To investigate whether the system could decode complex multi joint movement, the team used computer vision to track limb positions while recording from the motor cortex. They found that the μECoG array could decode coordinated movements of the paw, knee, and thigh in real time. Machine learning models identified which electrodes contributed most to each movement direction. Some electrodes were consistently useful for decoding multiple types of motion, while others were more specialized. This pattern reflects the spatial specificity of motor encoding in the brain and supports the idea that high resolution arrays are necessary to capture it.
The researchers also compared regions of the array that contributed more or less to decoding accuracy. The high contribution regions showed lower internal signal correlation, meaning they captured more distinct neural signals. These regions produced significantly better decoding results than neighboring electrodes of similar size and spacing. This finding reinforces the importance of spatial resolution for identifying small but information rich areas of the cortex involved in movement control.
To test real world feasibility, the researchers evaluated the system in human participants. In one case, a person undergoing awake brain surgery used the μECoG interface to play video games. After a brief calibration period using a joystick, the participant was able to control game elements using brain signals alone. In a one dimensional game, the person reached 90 percent accuracy. In a more complex two dimensional game, decoding accuracy reached 0.73 and 0.79 along each axis.
In a separate study, another participant used motor imagery to control a computer cursor across a 12 day trial. After initial adaptation, the person achieved full voluntary control using only imagined movement. In tasks requiring control of eight targets, the system reached a bit rate of 1.13. In more demanding tasks involving 255 targets, bit rates peaked at 4.15 following interface refinements. These rates match or exceed those reported in many studies of intracortical BCIs, suggesting that μECoG systems can offer competitive performance without requiring electrodes to penetrate brain tissue.
By combining long term stability with high spatial resolution and minimal invasiveness, the system developed by Zhou et al. offers a practical path forward for brain computer interfaces aimed at restoring movement. It provides consistent, high quality signals from the brain’s surface and supports real time decoding of complex motor activity in both animals and humans. The findings also clarify how array design affects performance, showing that density and spatial targeting matter more than overall coverage. These results point toward the use of compact, high precision μECoG interfaces for assistive technologies that are both clinically viable and adaptable to everyday use.
For authors and communications departmentsclick to open
Lay summary
Prefilled posts
ORCID information
Zhitao Zhou (University of Chinese Academy of Sciences)
, 0000-0003-4277-4461 corresponding author
Tiger H. Tao (Guangdong Institute of Intelligence Science and Technology)
, 0000-0002-6583-5039 corresponding author
Nanowerk Newsletter
Get our Nanotechnology Spotlight updates to your inbox!
Thank you!
You have successfully joined our subscriber list.
Become a Spotlight guest author! Join our large and growing group of guest contributors. Have you just published a scientific paper or have other exciting developments to share with the nanotechnology community? Here is how to publish on nanowerk.com.
