Construction of a Ten-Thousand-Channel Neural Graphics Card System for Invasive Visual Brain-Computer Interfaces

Science and Technology Daily News (Reporter Yang Xue): The reporter recently learned from XS Vision (Beijing) Technology Co., Ltd. (hereinafter referred to as “XS Vision”) that its independently developed invasive visual brain-computer interface has achieved breakthroughs in key areas such as advanced-process brain-computer chips with ten-thousand-channel capacity and clinical research on humans.

The fundamental principle of invasive visual brain-computer interfaces (BCIs) is to directly stimulate the visual cortex via high-density electrode arrays, encoding external image information into electrical signals that are “written” into the brain, thereby reconstructing artificial vision in cases where the optic nerve pathway is damaged. This technological approach is regarded as one of the most challenging frontiers in the field of visual rehabilitation. Its core challenge lies in enabling the brain not merely to “perceive light spots,” but to truly “see the world.”

Achieving ten-thousand-channel capacity is currently one of the key thresholds for measuring high-bandwidth, high-resolution information interaction capabilities in the international field of visual brain-computer interfaces (BCIs). XS Vision has constructed a neural graphics processing unit (GPU) system architecture with ten-thousand-channel capacity, dedicated to realizing large-scale neural signal acquisition and stimulation using electrode arrays with tens of thousands of channels. Currently, preliminary technical validation of this system has yielded initially positive feedback in supporting human clinical studies.

Meanwhile, the team successfully developed and validated a multi-scale neural interface system-on-chip (SoC), eliminating physical barriers to the full implantation of visual brain-computer interfaces and establishing a self-reliant and controllable foundation for China’s high-end brain-computer interface underlying computational hardware.

Clinical applications of brain-computer interface (BCI) technology must be grounded in authentic human brain data. It is reported that, leveraging over 100 cases of invasive electrophysiological and electrical stimulation data from the human visual cortex accumulated over the past decade, the team has established a robust data foundation. As a representative project for the commercialization of scientific and technological achievements at Peking University, XS Vision’s closed-loop model, combining “massive human data with proprietary high-throughput chips,” promises to pave the way for an independent and controllable technological pathway.