Breakthrough in Brain-Computer Interface Enables Precise Mind Control of Over 20 Digital and Physical Devices

Recently, the international authoritative journal Advanced Science published a significant research achievement jointly completed by Zhou Zhitao's team from the Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences, Tao Hu's team from NeuroXess, and Mao Ying/Chen Liang's team from Huashan Hospital, Fudan University—a universal implantable flexible brain-computer interface system with broad compatibility. The system, equipped with a brain-computer operating system, enables flexible brain control of various physical and digital devices, achieving precise mind control of more than 20 types of digital/physical devices. Under comparable training durations, its information transmission rate (BPS) matches the level of Neuralink’s subjects. Notably, this marks the first long-term implant clinical trial in China utilizing MEMS high-throughput, high-resolution flexible brain-computer interfaces, laying an important foundation for the clinical translation of high-throughput flexible brain-computer interfaces.

Innovation: Balancing Performance and Safety

For a long time, brain-computer interface technology has faced the dilemma of being unable to achieve both "high performance" and "high safety." The research team introduced that the existing mainstream technical routes each have limitations in signal acquisition methods — although electroencephalography (EEG) is non-invasive, because the signals need to pass through multiple layers of tissue such as the scalp and skull, they are severely attenuated, resulting in a low signal-to-noise ratio and limited spatiotemporal resolution, usually only allowing for the decoding of simple commands; intracranial EEG can achieve high-resolution signal acquisition at the single-neuron level, but it requires penetrating the cerebral cortex to implant electrodes, causing significant damage to brain tissue, easily triggering inflammation and immune responses, and having limited coverage; traditional electrocorticography (ECoG), while striking a balance between trauma and performance, has low electrode density and large equipment volume, not only limiting decoding accuracy but also often requiring extensive craniotomy, increasing the surgical risks and trauma for patients. To address these pain points, the research team innovatively adopted semiconductor micro-nano manufacturing processes to successfully develop an ultra-flexible, high-density 256-channel μECoG electrode array.

Reporters from Xinmin Evening News learned that this electrode has a density of 64 channels per square centimeter, which is 64 times higher than traditional ECoG electrodes. It also features excellent conformability – the ultra-thin mesh recording area can closely adhere to the cerebral cortex, ensuring high-fidelity signal acquisition, while the thickened lead area ensures long-term mechanical stability for implantation. The system is paired with a customized titanium alloy waterproof sealed casing and a low-power signal processing unit, ultimately achieving a triple technological breakthrough of "high throughput, high resolution, and low invasiveness."

Long-term: 203-day animal experiment verification

To verify the long-term safety of the system, the research team conducted a 203-day in vivo experiment on an 18-month-old Labrador weighing 30 kilograms. The experimental results showed that the μECoG electrode system exhibited excellent long-term stability: the signal frequency characteristics of the electrode channels remained consistent throughout the entire experimental period, with a signal-to-noise ratio consistently above 20dB, fully meeting real-time decoding requirements. In terms of motion decoding accuracy, the system maintained a decoding accuracy rate of over 78% for the position and velocity of the three-dimensional movements of the Labrador. The decoding accuracy in the Y direction (corresponding to knee flexion movement) reached up to 90%, with minimal fluctuations in decoding accuracy across all directions, demonstrating that the μECoG electrode system can stably capture neural signals related to fine motor movements.

More importantly, immunohistochemical analysis performed after the experiment showed that, compared with the contralateral homologous region of the brain, there was no significant neuronal loss in the electrode implantation area, and inflammatory markers such as astrocytes and microglia did not show a significant increase, fully verifying the long-term stability and biocompatibility of the system.

Clinical: Compatible with multiple scenarios of mind control

Based on animal experiments, the research team further conducted clinical validation. In one awake surgery for motor area localization, after only 7 minutes of model training, the patient was able to control brain activity through the μECoG electrode system to play ping-pong and snake games — with decoding accuracy for the ping-pong game (one-dimensional motion control) reaching 90%, and decoding accuracy for the X and Y directions in the snake game (two-dimensional direction and speed control) reaching 73% and 79%, respectively. This marks that the system can quickly adapt to the human body to achieve real-time motion decoding. In another clinical trial with an implant duration of less than one month, participants cumulatively completed 25,412 tasks (totaling 19.87 hours), covering task types such as Center-out and WebGrids paradigms. These tasks required participants to move the cursor to a highlighted target on the screen within 4 seconds and maintain it for 200 milliseconds; otherwise, the task was considered a failure. To help participants gradually adapt to cursor mind control based on motor imagery, the study adopted progressive training: flexible fixation algorithms were used to optimize the control experience for the first 6 days; from the 7th day onward, after about 30 minutes of calibration, participants could independently control the cursor using motor imagery without additional assistance, achieving a peak bitrate of 1.13 bits/second. For the more complex WebGrids task, after interface optimization, the peak bitrate further increased to 4.15 bits/second by the 9th day, comparable to the level of Neuralink's subjects. Ultimately, participants successfully achieved mind control over large complex games, smart wheelchairs, smart home devices, and various apps through the XessOs brain-computer operating system, fully demonstrating the broad clinical application prospects of this system.

The research team stated that the next step will be to continue optimizing system performance, actively promoting technology transfer, and accelerating the clinical implementation process. In addition, the team also looks forward to this system providing a powerful tool for fundamental research on neural coding and decoding mechanisms, further driving major breakthroughs in the field of brain science.

Original Title: A new breakthrough in brain-computer interface! Precise mind control of more than 20 digital/physical devices achieved