Can a thought-controlled wheelchair or robotic dog become as second nature to control as the movement of one’s legs and arms? In Shanghai, scientists have shown that this is no longer the stuff of science fiction but a reality in a clinical setting. The Centre for Excellence in Brain Science and Intelligence Technology (CEBSIT) of the Chinese Academy of Sciences developed a system that enabled a paralyzed patient to control several devices, such as a smart wheelchair, robotic dogs, and computer systems, with a fully implantable wireless brain computer interface (BCI) system for seamless interactions between the physical and digital realms.
The patient, Mr. Zhang, had a high spinal cord injury in 2022 that resulted in the loss of voluntary control over the rest of his body except in the area of the head and neck. After a year of traditional rehabilitation therapy without significant progress, surgeons at Huashan Hospital implanted the WRS01 BCI on June 20. This invasive BCI consists of highly flexible electrodes implanted in motor cortex areas that are connected to a processor located in the skull groove. The brain signals are then transmitted through an original wireless communication protocol to an external hat that provides inductive power. Within three weeks of training, Zhang was able to control the cursor of a computer through the BCI and began working as an intern sorter for the verification of AI recognition of vending machines at a distance marking the first recorded instance of someone earning a salary through the aid of BCI.
To accomplish the transition between lab trials and real-world integration, there was a need for the integration of different fields of engineering. Neural signal recording was improved by the use of high compression and high fidelity signal processing algorithms, incorporating spike band power, ISIs, and the number of spikes in a trial, to boost the accuracy of the signals by as much as 15–20%. Online “recalibration” algorithms enabled the parameters of the decoding to be adjusted in real-time, allowing for upgrading of the performance with practice and usage, rather than being handicapped by disuse between trials. More importantly, the latency between the thought and the onward action was reduced to “less than 100 ms,” with neural conduction taking 200 ms.
China has the advantage of being vertically integrated. This is made possible by high-bandwidth 5G or even 6G networks that facilitate fast, reliable communication between the implant and the device. Chip manufacturing companies provide miniaturized, biocompatible processors that use less power. Robotics companies provide adaptable models such as quadruped delivery robots and self-propelled wheelchairs. This combination drives fast translation from laboratory models to implementation.
Decoding signals involves the use of sophisticated AI algorithms that have learned to associate specific patterns of cortical activity and the corresponding desired tasks. WRS02’s updated system under the CEBSIT project will be extended to a total of 256 channels to capture more complex information for the decoding of outputs like syllables used in the Mandarin dialect. The high-density electrocorticography matrix has already resulted in a syllable accuracy level of 71.2% among the total 394 tonal syllables and has a sentence decoding rate of nearly 50 ch/min.
In addition to considerations in algorithms and hardware, asynchronous control, in other words self-paced control without predetermined trial intervals, is critical in practical applications. Methods such as correlation-based target selection with dynamic p-values for distinguishing control and non-control states can finally solve what is known as the ‘Midas Touch’ problem, in which unwanted commands might accidentally be selected during instances of idle thought. Such a function, developed in VEP-based spellers, is being transferred to motor-based BCIs for general applications.
Internationally, Elon Musk’s Neuralink has shown the ability of patients to play video games or control robotic limbs, but is still working with limited tasks. trials in China have already moved into multi-device, real-world independence. Pu Muming of the Chinese Academy of Sciences stressed the validation of electrode safety and stability, and successful decoding as immediate requirements before any medical applications. CEBSIT’s team chief, Zhao Zhengtuo, detailed plans: in three years, full restoration of motor and language abilities in the mainstream; in five years, restoration of perception with the development of prosthetic sight in the mainstream; and within a decade, the development of minimally invasive medical or consumer-grade enhancement products.
For the professionals in the fields of neuroscience and neuroengineering, it is evident that it is a case that shows the capabilities of the integration of various things like wireless communication, semiconductor development, AI-based signal processing, and robots in order to condense the time span from experimentation in the neuroprosthesis stage to achieving functional independence. China’s WRS01/WRS02 development in the competitive arena of applied neurotechnologies paves the way from Proof-of-Concepts to scalable and common Brain-Computer Interfaces and redefines what “mind control” will mean in reality.
