A village clinic in a remote county has no pathologist on staff. A woman comes in for a routine swab, and within a day she has a screened result on her phone — read not by a local doctor who does not exist, but by an algorithm in the cloud.
That is the quiet infrastructure China has been building for cervical-cancer screening, and the company at its center is Wuhan Lan Ding (武汉兰丁).
Key takeaways
- Lan Ding's AI digital pathology (数字病理) pipeline automates cervical-cancer (宫颈癌) screening from sample to report, with a doctor confirming suspicious cases.
- Its cloud system operates across 31 provincial-level regions and more than 2,000 partner medical institutions, with over 13 million women screened to date.
- The bottleneck it attacks is real: China has roughly one pathologist for every 140,000 people, and most specialists sit in top-tier city hospitals.
- Reported economics are striking — Lan Ding estimates one early-detected case can avert up to about ¥200,000 (≈ US$28,000) in later treatment costs.
- The model is now spreading beyond cervical cancer and beyond China, into Belt and Road countries.
The problem Lan Ding is solving
Pathology is the unglamorous backbone of cancer diagnosis, and China is short of it. State-media coverage of the national workforce cites roughly one pathologist per 140,000 people, with most concentrated in large urban hospitals. A county hospital often cannot staff a pathology department at all, which means women in rural areas either travel far for screening or skip it.
Cervical cancer is a textbook case for automation: it is common, slow-growing, and detectable from cell samples long before it becomes lethal. The catch has always been the human eye. Reading slides is tedious, and there are not enough trained readers.
How the pipeline works
Lan Ding, founded by Sun Xiaorong (孙小蓉), a pathologist who returned to China in 2000 after overseas study, built a full-process system rather than a single model. At a clinic, a sample is prepared on an automated staining machine, scanned by a smart scanner, and turned into a standardized data stream uploaded to the cloud. Lan Ding's vertical large model (垂直大模型) then performs a first-pass screen; suspicious cases are routed to a human pathologist for confirmation, and the patient can receive a report the same day.
The scale of the data is the differentiator. According to the company and state-media reporting, Lan Ding's cervical-cell dataset holds close to 13 million samples, totals around 1,400 TB, and feeds a model with about 1.5 billion parameters trained on more than one billion labeled images. The commercial system has run for roughly nine years and processes about two million cases a month.
Where it has landed
The network now spans 31 provincial-level regions and over 2,000 partner medical institutions, including township health centers. In Hubei, where Lan Ding is based, the system supported a provincial free-screening program from 2022 to 2024 and is credited by local officials with helping the province hit the World Health Organization's 70%-coverage target for eligible women about five years early.
Remote deployments tell the equity story best. In Shan'nan, Tibet — historically a poor, mountainous area with no pathology capacity — Lan Ding's AI gave local maternity hospitals a standardized, near-tertiary-level diagnostic ability after 2019. In September 2025, Henan opened its first "AI digital pathology diagnosis center" at Zhengzhou Yihe Hospital, designed to radiate services to surrounding county clinics.
The money and the mission
The economics are part of why this scales. Lan Ding says a single early-detected case can avert up to roughly ¥200,000 (≈ US$28,000) in downstream treatment costs, and the per-test cost is low enough for mass public-health programs. The company also notes it reinvests about 10% of revenue into research.
This is not only a Chinese story anymore. Lan Ding has partnered with more than ten Belt and Road countries and says it plans to work with fifty nations within five years, pitching a low-cost, cloud-based screening method as a replicable "China solution" for lower-resource health systems.
A wider trend, not a solo act
Lan Ding is the most deployment-heavy example, but it is not alone. In early 2025, Shanghai's Ruijin Hospital and Huawei Cloud released RuiPath, described as China's first clinical-grade pathology large model (大模型) to enter hospital production; reporting says it covers more than twenty hospitals and over 90% of common cancers with above-90% accuracy. The common thread: pathology, long bottlenecked by scarce specialists, is becoming the front line where Chinese AI-in-practice meets public health.
Honest limitations
Most figures here — screening counts, dataset size, cost savings, institutional reach — come from the company and from Chinese state-media reporting, not from independently audited disclosures. AI screening is a triage and first-read step; a human pathologist still confirms positives, and the system should not be read as replacing diagnosis. Generalization from cervical cells to other cancers (oral, thyroid, bladder) is underway but not yet as proven at scale. The 1-in-140,000 pathologist ratio is a widely cited media figure, not a precise census. I have not visited a Lan Ding deployment, and cross-border rollouts depend on local regulatory acceptance.
What readers can do now
- If you are in a screening-eligible group in China, ask whether your local clinic uses AI-driven (云诊断) pathology triage — it can shorten wait times and catch cases earlier.
- Health-system watchers elsewhere should track Lan Ding and RuiPath as case studies in cloud-based pathology for underserved regions, not just as products.
- Anyone evaluating "AI for healthcare" claims should insist on the confirmation step: the useful deployment is human-in-the-loop triage, not autonomous diagnosis.