--- title: "A Wuhan startup built a 65,536-channel brain implant — and took the idea into surgery" date: 2026-11-09 category: Brain–Computer Interface site: NeuroAI canonical: https://neuroai.site/a/na-brain-zhonghua-65536-channel-bci language: en --- # A Wuhan startup built a 65,536-channel brain implant — and took the idea into surgery > Wuhan Zhonghua's 65,536-channel invasive brain–computer interface cleared its first all-domestic human test in late 2024, a high-density Chinese challenge to penetrating arrays. A surgeon in Wuhan lowered a postage-stamp-sized device onto the surface of a patient's brain and watched it read signals no scalpel could see. The patient was awake enough to cooperate, and the implant was not imported, not licensed, not a copy. It was designed and built entirely in China. That device belongs to Wuhan Zhonghua BCI (武汉衷华脑机), a company that has spent the last few years quietly assembling one of the highest-channel-count invasive brain–computer interface (脑机接口) systems ever reported. ## Key takeaways - Zhonghua's flagship implant carries 65,536 channels — built from 1,024 flexible "soft needles," each holding 64 contacts — far above the roughly 1,000-channel arrays most human implants use today. - In November 2024 the system completed its first all-domestic, high-channel clinical test at Wuhan Union Hospital (Xiehe) on a brain-tumor patient, used to map tumor margins via neural signals. - The company says it has filed more than 200 patents and plans staged clinical trials across conditions including paralysis, stroke, Parkinson's, epilepsy and depression. ## What 65,536 channels actually means Most invasive brain implants record from on the order of a thousand sites. Zhonghua's array multiplies that by arranging 1,024 thin, flexible needles, each tipped with 64 electrodes, for a total of 65,536 recording points. The pitch from the company is density plus reversibility: the needles are soft enough to contact brain tissue without the rigid penetration that worries clinicians, and the system is described as bidirectional, able to both read neural activity and write stimulation back. Whether that density translates into meaningfully better decoding is the open question every high-channel claim faces. More points can mean cleaner spatial maps of brain activity, but also more signal to manage, more power, and more risk of information that looks rich but decodes poorly. Zhonghua has not yet published head-to-head accuracy data against lower-channel systems. ## The human test that mattered The headline event was not a laboratory demo but a hospital operating room. In November 2024, Zhonghua and Wuhan Union Hospital reported the first fully homegrown high-channel brain–computer interface clinical test in China. The subject was a brain-tumor patient; surgeons used the implant's neural readout to help distinguish tumor tissue from healthy brain, a task where margins are notoriously hard to see. That is a recording task, not a stimulation one, so the case exercised the read side of a system the company describes as bidirectional. The device in that operation was the company's flexible epidural array, laid on the brain's outer covering rather than driven into tissue; the 65,536-channel microneedle figure describes the flagship specification, and the company has not published how many channels that first implant actually carried. Before that, in August 2023, an expert panel led by academician Zhao Jizong evaluated the "65,000-channel bidirectional implantable BCI system" and judged it internationally leading. The company says it has since run several short-term clinical trials across Union and Tongji hospitals and is preparing longer-term implants. ## Why a real-estate fortune is behind it Zhonghua was founded in 2021 and is chaired by Huang Li, better known as the founder of infrared-sensor maker Guide Infrared. The BCI effort drew on more than a decade of MEMS chip experience and was, by the company's account, self-funded through its early years. That background matters: building a 65,536-channel system is less a software trick than a manufacturing problem — fabricating, aligning and packaging tens of thousands of contacts reliably enough to put near a human brain. ## The China context Zhonghua is one of several Chinese teams racing up the channel-count ladder, sitting alongside invasive programs from Tsinghua, Fudan and commercial players. Its bet is that domestic, end-to-end control of the MEMS process, the chip and the algorithm is the real moat — not any single spec sheet. If the longer-term trials proceed, the company says it aims for a Class III medical-device production license within a couple of years. ## What readers can do now - **Treat channel count as a means, not a verdict**: ask what a high-density implant actually lets a patient do, not just how many electrodes it has. - **Track the clinical registry, not the press release**: real signal will appear when Zhonghua publishes trial endpoints, adverse events and decoding accuracy. - **Compare architectures fairly**: surface films, vascular stents and penetrating needles each trade signal quality against surgical risk — none is simply "best." ## Honest limitations This article is based on Zhonghua's corporate communications, Wuhan municipal and Hubei provincial government portals, and Chinese media reporting of the November 2024 clinical test; we did not access peer-reviewed trial data or raw surgical records. The 65,536-channel figure and the "first all-domestic high-channel clinical test" claim come from the company and affiliated hospital announcements and have not been independently re-audited here. The company's comparison of its density to foreign peers is presented as its own characterization; we avoided repeating a specific multiplier because channel-count comparisons depend on what is being counted. Long-term safety, decoding performance across patients, and the timeline to a production license remain unverified pending public clinical evidence. --- Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-brain-zhonghua-65536-channel-bci Free to quote with attribution and a link to the original.