China’s semi-invasive BMI enables patient to speak Chinese

China’s semi-invasive BMI enables patient to speak Chinese and allow paralyzed individuals to control robotic arms.

A Chinese research team specializing in semi-invasive brain-machine interface (BMI) technology has successfully conducted multiple full cortical BMI implantations, enabling an aphasic patient to communicate in Chinese and allowing paralyzed individuals to control computers and robotic arms.

At a press conference held on Thursday by the Chinese Institute for Brain Research, Beijing (CIBR), Xuanwu Hospital (affiliated with Capital Medical University), and NeuCyber NeuroTech (Beijing) Co., Ltd., it was revealed that over 98 percent of BMI channels remained functional after implantation, as demonstrated in the first group of patients.

The NeuCyber Matrix BMI System, jointly developed by CIBR and NeuCyber, is a wireless semi-invasive device featuring a flexible, ultra-thin nano-fabricated film microelectrode. It consists of a 128-channel interface integrated with a compact micro-circuit that records and processes electrocorticography (ECoG) signals.

For the first time, an aphasic patient has successfully communicated in Chinese using the semi-invasive BMI system, restoring their ability to express themselves. Paralyzed patients have also adapted to the system, using it to operate external devices and regain some motor functions.

CIBR Director Luo Minmin explained that the system deciphers ECoG signals through three core technologies: a highly integrated micro-host that processes large-scale, low-power signals; advanced short-range wireless communication for high-bandwidth data transfer; and a real-time, multi-scenario algorithm capable of accurately decoding Chinese language and fine movements.

Earlier this month, a surgical team led by Xuanwu Hospital President Zhao Guoguang performed the world’s first wireless BMI implantation for the Chinese language on an ALS patient who had lost the ability to speak.

Using a neurosurgical robotic system, the team implanted the NeuCyber Matrix BMI System onto the dura mater on the left side of the brain, which is responsible for language processing. Unlike fully invasive procedures that implant electrodes inside brain tissue, this semi-invasive approach places electrodes externally, reducing trauma while capturing high-quality neural signals.

A small, coin-sized control and signal transmission device was implanted on the patient’s skull, allowing wireless transmission of neural signals and power supply through near-field communication technology.

Following the surgery, the patient began language decoding training on March 14. Within three hours, the real-time accuracy of decoding 62 commonly used words reached 34 percent, later improving to 52 percent.

With the assistance of an adaptive error correction algorithm based on a large language model, the patient has regained basic communication abilities. The system has successfully decoded sentences such as “I want to drink water,” “I want to eat,” and “I’m in a great mood today. I’d like to take a walk with my family.” Currently, the decoding latency for a single Chinese character is under 100 milliseconds.

Li Yuan, business development director at NeuCyber NeuroTech, highlighted that BMI technology enables speech-impaired patients to regain communication, and with synthetic speech integration, their thoughts can even be vocalized.

Zhao emphasized that the success of this semi-invasive BMI technology presents a promising long-term solution for aphasic patients, expanding its potential applications in diagnosing and treating neurological disorders. His team plans to further explore BMI applications for conditions such as epilepsy, spinal cord injuries, stroke, ALS, and aphasia.