Decoding the "black technology" of brain-computer interfaces: How far has the breakthrough in artificial vision come?

How to generate higher-quality images remains the core challenge facing scientists.

Recently, Mingshi Brain-Machine Interface, a rising star in the domestic brain-computer interface field, announced that it has achieved the world's first functional interactive verification of visual reconstruction for "complex graphics + multiple colors".

According to a reporter from the Science and Technology Innovation Board Daily, similar to the reconstruction of motor and language functions, the repair and reconstruction of vision is another crucial direction for brain-computer interface technology in the medical field. Its history can be traced back to the 1990s, but it still faces an insurmountable technological gap, namely how to achieve the leap from simple light spots to complex graphics and even color perception in visual perception through electrical stimulation of the visual cortex of the brain.

"The significance of Mingshi Brain-Machine Interface's current research lies in the dynamic analysis and reconstruction of graphic outlines and basic colors. This marks that China has entered the world's first echelon in this field and proposes a highly promising technological solution. However, whether it will ultimately succeed still needs to be verified by future public, rigorous scientific data and large-scale clinical trials," Dr. Liu Bing, the founder and CEO of Mingshi Brain-Machine Interface, told the Science and Technology Innovation Board Daily reporter.

From Retinal Prostheses to Visual Cortex Prostheses

When light enters the eye, it first passes through the cornea and lens, the outer and middle layers of the eye, and then reaches the retina at the back of the eye. The photoreceptor cells in the retina convert the light signals into biological current pulses, which are transmitted to the visual center of the brain via the optic nerve. After complex neural processing, the clear visual images we perceive are finally formed.

This is the process of human vision. Any problem in this process may lead to visual impairment or even blindness. For example, cataracts can make the lens cloudy, blocking the normal passage of light; some retinal diseases such as age-related macular degeneration and diabetic retinopathy can damage or kill photoreceptor cells, preventing them from effectively receiving light signals; and conditions such as glaucoma, trauma, or tumor compression may damage the optic nerve or the visual center of the brain, resulting in blocked signal transmission.

To help blind people "see the light again", scientists have developed visual prostheses that can provide "artificial vision" for the blind. According to the implantation site, visual prostheses are mainly divided into two categories: one is the retinal prosthesis that focuses on intraocular repair, and the other is the visual cortex prosthesis that tries to bypass the eye and directly act on the visual cortex of the brain.

Among them, retinal prostheses are currently more widely used. Its core principle is to replace or repair the photoreceptor cells in the retina that have lost their function through an artificial device, thereby helping patients restore some visual perception. Globally, several landmark products and companies have emerged in this field. For example, the Argus II epiretinal prosthesis developed by Second Sight in the United States was approved by the US FDA in 2013, becoming the first retinal prosthesis to receive FDA approval. In the same year, the Alpha-IMS subretinal prosthesis developed by Retina Implant AG in Germany also received authorization from the European EMA and was allowed to enter the market.

However, due to various constraints, the commercialization of these products after their launch was not smooth, and they gradually faded out of the market. On the one hand, the high costs of R & D, production, surgical implantation, and postoperative maintenance greatly limited the popularization of the products; on the other hand, the early retinal prosthesis products were still in their infancy in terms of technology, and the visual resolution they could provide was limited, far from natural vision, making it difficult to truly meet the actual needs of patients.

Recently, at the "2025 Brain-Computer Interface Conference" held in Shanghai, Yu Xinguang, the director of the Brain Science and Neurology Center of Guangdong Heyou International Hospital, also systematically summarized the deficiencies and limitations of artificial retinal prosthesis technology.

He pointed out that among several visual repair solutions such as retinal prostheses, optic nerve prostheses, and visual cortex prostheses, retinal prostheses are applicable to the most limited patient group, mainly for patients who are blind due to retinal degeneration, accounting for only about 10% of all blind people. In contrast, visual cortex prostheses are almost applicable to all totally blind patients for implantation, with the widest scope of application.

"Meanwhile, the field of view of these retinal prostheses is usually relatively narrow. For example, the field of view of the Argus II is less than 30 degrees. Users have to frequently turn their heads to scan the surrounding scene to compensate for the limited field of view. But this not only affects visual perception and the efficiency of daily activities but also easily causes discomfort over time," Yu Xinguang said.

Against this background, the industry has gradually shifted its research focus from retinal prostheses to visual cortex prostheses. As early as May 2019, Second Sight announced that it would stop producing the Argus II and fully shift its R & D resources to the development of the Orion visual cortex prosthesis system.

Breakthroughs and Limitations of Artificial Vision

Visual cortex prostheses are the cutting-edge direction in the current field of visual repair. Its core principle is that the visual impairments of many blind people are mainly due to damage to the eyes or optic nerves, but the areas in their brains that process vision are often intact. Visual cortex prostheses can bypass the damaged eyes and optic nerves and directly transmit visual information to the brain, thereby helping blind people restore some visual functions.

Neuralink, founded by Elon Musk, has focused on this direction. Their visual restoration project called "Blindsight" has not only received the Breakthrough Device Designation from the US FDA but also plans to enter clinical trials around 2026.

It should be noted that although the principle is clear, the R & D of visual cortex prostheses still needs to overcome a series of huge technological obstacles. Among many challenges, how to generate higher-quality images is arguably the most core problem.

As Philip Troyk, a professor of biomedical engineering at the Illinois Institute of Technology, emphasized, "The current work in this area (R & D of visual cortex prostheses) is not aimed at restoring biological vision but exploring the possibility of artificial vision." In other words, the artificial vision that the current technology can provide to users is still very limited.

Liu Bing told the Science and Technology Innovation Board Daily reporter that the effects of traditional visual repair technologies, whether retinal prostheses or early cortical stimulation visual prostheses, mostly stay at allowing users to perceive isolated 'light spots', which are scientifically known as phosphenes. This is like being able to light up only a few pixels on a screen but unable to form a meaningful image.

Currently, scientists are still seeking solutions. Research directions include increasing the density of electrodes to improve spatial resolution, optimizing the implantation position of electrodes, and innovating stimulation strategies. For example, some believe that more electrodes are not necessarily better. The key lies in the implantation position. If the electrodes are distributed in multiple areas of the visual cortex, it may stimulate more light spots in a larger field of view.

In addition, innovation in stimulation patterns is also important. Yu Xinguang gave an example, saying, "Existing research has shown that if the electrodes are stimulated in a specific order according to the writing stroke sequence, the human brain can better recognize the shape of letters."

This time, Mingshi Brain-Machine Interface successfully completed the functional interactive verification of visual reconstruction for "complex graphics + multiple colors". According to the Science and Technology Innovation Board Daily reporter, its core lies in adopting a closed-loop adaptive path of 'brain-machine dual learning' instead of the traditional open-loop, fixed stimulation path.

The traditional path is like "one-way indoctrination". A set of stimulation patterns is preset first, and the brain is made to adapt to the machine. However, the brain itself is dynamically changing, and fixed algorithms often lead to a rapid decline in effectiveness.

"In contrast, our path is a 'two-way dialogue'," Liu Bing explained to the Science and Technology Innovation Board Daily reporter. "The system can not only precisely stimulate and encode the visual cortex through a high-density electrode array but also read the neural feedback from the brain in real-time. The stimulation strategy and decoding model can be dynamically optimized, thereby achieving common adaptation and learning with the user's brain. This fundamentally solves the industry problem of long-term stability."

Regarding the principle of achieving color perception, he further said, "We simulate the neural activity patterns triggered by light of different wavelengths in specific functional areas of the visual cortex through specific electrical stimulation sequences. When the brain receives such 'simulated' signals, it will interpret the corresponding colors."

"Currently, we have stably achieved the distinction and perception of basic colors such as red, green, and blue, proving the feasibility of encoding and transmitting color information through cortical electrical stimulation. This is the first step from the black-and-white world to the colorful world. Of course, there is still a long way to go to restore all the colors and delicate tones in nature, but the door has been opened," Liu Bing said.

He compared the current progress to "creating the first 'display prototype' that can show graphics and colors" and said that the Mingshi Brain-Machine Interface team will significantly improve the 'display resolution' by increasing the electrode density, optimizing the encoding algorithm, and integrating key technologies such as computer vision.

This article is from the WeChat official account "Science and Technology Innovation Board Daily". Author: Xu Hong. Republished with permission from 36Kr.