Major Breakthrough in Brain-Computer Interface: Mind-Controlled 'Thought-to-Speech' Becomes Reality

At the beginning of the new year,重磅 news came from the brain-computer interface field.

On January 2, NeuroXess, a leading company in invasive brain-computer interface technology in China, announced a breakthrough in human clinical trials. In close collaboration with the Department of Neurosurgery at Huashan Hospital affiliated with Fudan University, and with support from the Tianqiao Institute of Brain Science, the company achieved significant progress. Based on its fully self-developed 256-channel high-throughput implantable flexible brain-computer interface, the team conducted clinical trial research on high-precision real-time motion decoding and language decoding, successfully enabling "brain-controlled" smart devices and "mind-based communication."

This breakthrough not only covers real-time motion decoding but also achieves an unprecedented leap in real-time Chinese language decoding, marking China's rise to a world-leading position in the brain-computer interface field.

The entire brain-computer interface technology has undergone a journey from academia to scientific research, from scientific research to industry, and from industry to commercialization. It is now entering a phase of rapid commercialization.

Neuralink, founded by Musk, has released a series of latest achievements in the past two years, pushing the industry to evolve more rapidly, and the influx of capital has also become more aggressive.

　　East MoneySecurities analysis points out that the current market size of brain-computer interface equipment in China has reached billions, and is expected to reach hundreds of billions by 2040. China's policies strongly support the development of brain science and brain-inspired research, with close collaboration between industry, academia, research, and medicine driving technological progress. The large patient population base expands demand, while leading companies such as NeuroXess propel the industry forward. With multiple contributing factors, the brain-computer interface sector is poised to enter a fast track of development.

More notably, brain-computer interfaces are expected to be combined with artificial intelligence technology, and together they will push human society into a new era of intelligence. This involves not only technological advancements but also the enhancement and expansion of human capabilities.

NeuroXess Founder and CEO Peng Lei told the 21st Century Business Herald that he believes the integration of silicon-based life and carbon-based life will occur between 2035 and 2045. "We represent the trend of carbon-based life embracing silicon-based life, while other AI partners represent silicon-based life simulating our ways. Anyway, sooner or later, we will meet in the middle."

Dual Decoding of Motor Language

On January 28, 2024, Neuralink, the brain-computer interface company founded by Elon Musk, successfully performed the first human brain chip implant surgery.

In March, Musk revealed on social media that this patient, who had suffered a spinal cord injury due to a diving accident 8 years ago and became quadriplegic, was now able to play games by controlling the mouse with his mind.

This marks the transition of brain-computer interface technology from theoretical research to practical application, with the potential to achieve significant breakthroughs in commercial applications and medical treatments.

Relevant experiments in China are also underway. In August 2024, NeuroXess, in collaboration with the neurosurgery team of Professor Mao Ying and Chen Liang from Huashan Hospital, successfully completed a clinical trial on motor imagery synthesis. The subject was a 21-year-old epilepsy patient with a lesion in the motor cortex. A 256-channel high-density flexible brain-computer interface was surgically implanted to monitor the lesion and protect critical motor-related brain functional areas.

The High Gamma band (70-150Hz) is a high-frequency band in EEG signals, typically associated with complex cognitive functions and neural activity synchronization in the brain. It provides detailed information on brain activity, especially motor and sensory information. In clinical trials of implantable brain-computer interfaces, extracting signals from this frequency band helps decode brain intentions more accurately, enabling thought-driven synthetic motion.

According to the introduction, the project team extracted EEG features and conducted model training on the High Gamma band of the brain's electrical signals. They utilized an LSTM (Long Short-Term Memory) neural network model for continuous time decoding, with the overall system latency being less than 60ms. Thanks to the 256-channel high-throughput, high-quality, and high-resolution EEG signals, combined with a self-developed channel selection algorithm, the responsive brain regions can be quickly and accurately identified, enabling real-time, efficient, and precise decoding.

Subjects are not required to operate manually and can achieve "brain-controlled" play of ping-pong and Snake games within two days after the surgery. Notably, in addition to achieving real-time motion decoding, the breakthrough of NeuroXess lies in accomplishing real-time Chinese decoding—currently, Neuralink, founded by Elon Musk, has only achieved real-time motion decoding.

It is widely believed in the industry that decoding language is the next major breakthrough for brain-computer interfaces. Language involves the integration of various brain functions such as cognition, memory, and emotion, making its complexity far surpass simple motor control. Currently, there are only a few scientists worldwide working on decoding language through brain-computer interfaces, and most of them focus on English.

Compared with alphabetic languages such as English, Chinese decoding faces higher technical difficulty. Chinese is a tonal and graphic language dominated by monosyllables, which differs from alphabetic languages like English. The information conversion in its production process involves more brain regions. This requires brain-computer interface technology to cover all relevant brain areas and collect sufficient data to accurately "read" Chinese information in electroencephalograms, necessitating the development of neural encoding and decoding mechanisms and information processing methods tailored to the characteristics of the Chinese language.

In December 2024, NeuroXess, in collaboration with Professor Wu Jinsong's team from the Department of Neurosurgery at Huashan Hospital, conducted the first clinical trial in China for real-time synthesis of the Chinese language using a high-throughput implantable flexible brain-computer interface. The project team performed a flexible brain-computer interface implant surgery on a patient with epilepsy caused by a tumor occupying the language area, using a 256-channel high-throughput brain-computer interface to help locate the lesion and protect important brain functional areas related to language.

According to the project team, the patient recovered well after the surgery, achieving a 71% decoding accuracy rate for 142 commonly used Chinese syllables within five days. Additionally, the latency for single-character decoding was less than 100ms, representing the highest level of real-time Chinese language decoding currently in China.

This experiment successfully translated patients' thoughts into language, which was then converted into corresponding instructions. For patients with motor and speech impairments caused by brain injuries, such as those with aphasia, this technology holds promise in helping them restore their language abilities and communicate with the outside world. More importantly, this breakthrough also opens up new possibilities for direct interaction between the human brain and large AI models, or even thought-based communication, which will have profound implications for future communication methods and human-computer interaction.

As Tao Hu, founder and chief scientist of NeuroXess, previously stated at the Artificial Intelligence Conference, brain-computer interface companies focus on two main goals: one is to restore patients’ health by conducting extensive clinical research and targeting treatments for major brain diseases, aiming to return patients to a normal state; the other is to explore the technical limits, investigating the ceiling and imaginative boundaries of brain-computer interface technology by integrating the brain with more powerful sensors and actuators, extending the functions of the five senses and limbs.

Brain-Computer Interface and AI Complement Each Other

Currently, intelligent technology centered on Artificial General Intelligence (AGI) is advancing rapidly, showing an unpredictable development pattern characterized by holistic mutations and a surge of new phenomena. This, in turn, drives related industries to innovate and self-evolve. Given the powerful impetus that future brain-computer interface technology can bring to social development, it will inevitably couple with artificial intelligence technology, giving rise to new productive forces and momentum for development, thereby empowering humanity.

Peng Lei pointed out at the previous Artificial Intelligence Conference: "Brain-computer interface and AI are like two sides of a bridge, and we are developing towards each other."

Currently, the AI field is training models with hundreds of billions of parameters to mimic the human brain's functioning as closely as possible. At the same time, brain-computer interface technology is dedicated to converting brain activity into data and algorithms that computers can understand, achieving direct communication between the human brain and computers through methods like electrode implantation. This process of mutual integration suggests that carbon-based life (biological organisms) and silicon-based life (computer systems) may become more closely connected in the future.

"Although it is currently unclear which side will develop faster, their development direction is the same," said Peng Lei, referring to achieving deeper human-computer interaction and intelligent applications.

With the collaboration and joint exploration between brain science and AI, it is expected that these two fields will achieve integration more rapidly, thereby propelling human society into a new era of intelligence. Therefore, building a bridge between brain-computer interface technology and AI technology to achieve their mutual reinforcement and common progress has become a core issue in the current industry.

However, the development speed of brain-computer interfaces is slower than AI in some aspects and still faces some challenges and problems. Currently, the brain-computer interface industry lacks a mature industrial chain, and industry standards need further improvement.

Specifically, the hardware level is not unified, with different research institutions using different devices, electrodes, and chips; the software level is not unified, as researchers may use self-developed software or software provided by different companies, connecting through custom-made interfaces; data is not unified, with a lack of sharing between different studies, which limits the accumulation and validation of knowledge; algorithms are not unified, with research results often being kept closed, failing to form an effective open ecosystem, hindering further technological development and application.

In contrast, the rapid development of the artificial intelligence field has benefited from the standardization and unification of its foundational capabilities, laying the groundwork for the swift iteration and widespread application of AI technologies.

In this regard, creating a unified open platform for brain science is particularly necessary. Peng Lei stated that, on the one hand, data sharing and algorithm openness must be achieved by making hardware, software, data, and algorithm resources available to scientists worldwide. This will promote cumulative data, shareable algorithms, and accelerate the iteration of the entire ecosystem. On the other hand, more collaboration opportunities are needed, such as working with cross-border scientists, animal research institutions, and experimental platforms to find as many unified models as possible for similar research, forming a virtuous closed-loop ecosystem with shared resources and accessible information.

"Whether scientists, doctors, scholars, or industry partners, our goal is to work together to build an ecosystem that empowers brain science to advance at the iterative speed of AI, just as AI transforms various industries," Peng Lei revealed, adding that he plans to further implement the Brain Science Open Platform.

Brain-computer interface technology is undoubtedly a cutting-edge science with revolutionary potential, heralding a new direction for human-computer interaction. Despite challenges in technical implementation, ethical considerations, and safety assessments, industry professionals generally hold an optimistic view on the commercial prospects of brain-computer interfaces. With the rapid advancement of technology and continuous deepening of research, brain-computer interface technology will also find increasingly widespread applications.

Zhou Mingzi, Executive Director of Frost & Sullivan Greater China, believes that under the high attention of China's top-level design, effectively promoting the technological progress and innovation of brain-computer interface, as well as steadily and standardly advancing the industrial development of brain-computer interface are top priorities. Currently, China’s brain-computer interface industry lacks a mature industrial chain, and industry standards need further improvement.

Editor: He Songlin