A screening tool that started as a conference demo is now quietly reading scans inside hundreds of Chinese hospitals. Most patients never see it, and that is exactly the point. The question worth asking is not whether AI can read an image, but how a country actually folds it into clinical routine.
From one demo to a regulated product
Tencent launched Miying (觅影) on 3 August 2017, pitching it as the company's first AI product aimed at medicine. Early versions targeted early screening for esophageal and lung disease using image recognition and deep learning. Talk was cheap in 2017; what mattered later was clearance.
In August 2021, Miying's "pneumonia CT imaging assistant triage and assessment software" received a Class III medical-device registration certificate from the National Medical Products Administration (NMPA, 国家药监局) — registration number 国械注准20213210612. That made Tencent the first Chinese internet company to hold a Class III certificate for a medical AI product. Class III is the strictest tier in China, reserved for devices where safety and efficacy are hardest to prove.
Since then the certificate count has widened. Public reports put Miying at six medical-device registrations spanning respiratory, ophthalmology and digestive disease. A colorectal-polyp endoscopy detection product cleared review as an innovative device and was approved on 1 June 2023.
What it actually does in the clinic
Miying is not one model but a stack of disease-specific detectors:
- Lung and chest imaging — flags suspicious nodules and, from the pandemic onward, helped triage pneumonia CT scans.
- Fundus (eye) screening — in a few seconds analyzes retinal images for chronic glaucoma-like optic neuropathy and diabetic retinopathy.
- Digestive endoscopy — highlights colorectal polyps during a procedure to lift detection rates.
- A cloud layer — the "imaging cloud" lets hospitals share data and run remote diagnosis across campuses.
In 2024, Tencent Health used Miying's fundus technology for free public screening in places like Shenzhen and Ganzhou, Jiangxi, tallying more than 200,000 screening sessions.
The newer push wraps these detectors in a large model (大模型) layer — Tencent's Hunyuan (混元) foundation model — so the system can also interpret reports and support whole-episode patient management, not just spot a lesion.
Why 500 hospitals matters
By September 2025, Miying's imaging AI had reached more than 500 medical institutions across China and cumulatively assisted clinicians on close to 10 million patient exams, according to Tencent's own healthcare division. Those are deployment numbers, not benchmark scores, and they signal something the industry struggled with for years: crossing from lab to ward.
Two things helped. First, Miying was named in November 2017 by the Ministry of Science and Technology as one of the first National New Generation AI Open Innovation Platforms, specifically for medical imaging — a policy tailwind that opened hospital doors. Second, the company built an open lab so doctors and hospitals could train and validate their own models on Tencent's cloud, lowering the barrier for smaller institutions.
The honest part about "assist"
Miying is built to assist, not to decide. Tencent Health's own line is that the doctor keeps the final call on every diagnosis and treatment plan. That framing is not just courtesy — in a regulated medical market, the device is a registered aid, and the physician signs the report.
What readers can do now
- If you run a clinic or hospital system in Asia, treat NMPA Class III clearance as the real bar when evaluating Chinese imaging AI — Miying's certificate trail is the template to compare against.
- If you build health AI, study the "open lab + cloud" distribution model; regulatory approval alone did not drive adoption, hospital-grade tooling did.
- If you report on AI in medicine, track deployment counts (institutions, exams) over demo videos — they are the harder number to fake.
Honest limitations
Core facts: Miying launched 3 August 2017; NMPA Class III certificate for pneumonia CT triage software in August 2021 (registration 国械注准20213210612), first Chinese internet company to hold one; six medical-device registrations; colorectal-polyp endoscopy product approved 1 June 2023; more than 500 institutions and close to 10 million assisted exams as of September 2025; designated a National New Generation AI Open Innovation Platform for medical imaging in November 2017; 200,000+ free fundus screenings in 2024. These come from Tencent's official healthcare newsroom (21 May 2026 and 5 June 2026 statements by Tencent Health president Wu Wendai), People's Daily/IT之家, Nanfang Daily (NFNews), and Baidu Baike's compiled timeline. The "nearly 10 million exams" and "500+ hospitals" figures are Tencent's own disclosed deployment numbers, not an independent audit. The "large model (大模型)" and "Hunyuan (混元)" references describe Tencent's current product direction as stated by the company. This article contains no RMB amounts, so no currency conversion applies. Analysis is current to 29 September 2026 and is not medical or investment advice.