A farmer with chest pain arrives at the county hospital in Nanpi, Hebei — a two-hour drive from Beijing's tertiary centers. The CT scanner is mid-range, the radiology team is small, and until recently a chest report took the better part of a working break to write. Now, according to the hospital, it takes about three minutes, with an AI system drafting the first pass.
The county-level gap China is trying to close
China's best medical AI has concentrated where the patients and budgets are: Class-III tertiary hospitals in big cities. The policy pressure runs the other way. Under the "county medical community" (县域医共体) model, a county hospital is supposed to anchor and upskill every township clinic beneath it — usually with thin staffing and tighter budgets than any urban flagship.
Nanpi County People's Hospital became one test case. In 2025 it deployed a full-chain AI system from Shukun Technology (数坤科技), a Beijing medical-imaging company, covering record generation, clinical decision support, quality control, imaging and ultrasound. The hospital presented results at the 2026 Rural Health Development Conference in Chengdu in May, where its approach was singled out by experts affiliated with the National Health Commission's county-community expert group.
What the AI actually does there
Per the hospital's presentation and Shukun's account of the deployment:
- Chest CT reporting fell from 8–10 minutes to roughly 3 minutes — an efficiency gain the hospital puts at more than 30% — with the AI also flagging tiny nodules, bone-density loss and coronary calcium in the same scan
- Smart medical records generated from voice input and linked test data are in routine use in more than half of the hospital's departments
- Real-time quality control checks records for contradictions — mismatched sex or anatomy codes, medication-surgery conflicts — before filing
- AI-enabled portable ultrasound, combining Shukun's software with a Mindray handheld device, has been delivered to the county hospital and extended to township-level facilities, letting less-experienced operators get scan guidance on screen
The county says the program helped it win support under a national primary-care strengthening project (强基工程). The claims come from the company and hospital materials, not an independent audit — but they describe deployment details, workflows and conference appearances that are specific enough to be checked.
The company behind it
Shukun was founded in Beijing in 2017 and sells a "Digital Doctor" (数字医生) line of more than 40 AI products spanning heart, brain, chest, abdomen and musculoskeletal imaging. According to company disclosures cited by industry trackers, as of March 2025 it held 17 NMPA Class-III registrations — the strictest Chinese device class — plus 14 Class-II registrations, 10 EU MDR CE marks, one Japanese PMDA approval and two FDA clearances, with products in more than 2,000 Chinese hospitals (some trackers put the figure near 3,000).
Its 2026 has been an export push: in January, its coronary CTA and lung-nodule AI won Japanese PMDA certification, and in June two newer AI products — a liver-MR assistant and a gated calcium-scoring tool — received EU MDR CE marks, opening European market access.
Why this case matters more than the numbers
The Nanpi story is less about any single efficiency percentage and more about the deployment pattern: one AI vendor, one county, every workflow from registration to ultrasound to quality control. If that pattern replicates, China's medical-AI market shifts from selling point solutions to elite hospitals toward full-stack contracts with thousands of county hospitals — a bigger, if lower-margin, market, and one with a policy tailwind behind it.
Honest limitations
- The Nanpi figures (report time, 30% efficiency gain, department coverage) come from Shukun's own news release and the hospital's conference presentation; no independent audit or peer-reviewed study of this specific deployment is public.
- "More than 30%" is the hospital's own characterization of reporting-time gains; baseline workloads and case mix were not disclosed.
- Regulatory counts are as of the dates cited and change as approvals accumulate; hospital-count figures vary between the company site (2,000+) and third-party trackers (nearly 3,000).
- Whether the Nanpi model is financially sustainable without government project funding is not addressed in the available materials.
Sources for this article are Shukun Technology's official news releases (including the Nanpi deployment account and 2026 PMDA/CE certification announcements), the hospital's presentation at the 2026 Rural Health Development Conference as reported by Shukun, and independent industry trackers (including regulatory-count summaries) for company background. Company-published claims are attributed throughout and were not independently verified. Current to 2 October 2026.
What readers can do now
- If you procure clinical AI, copy the county checklist: measure report-turnaround before and after, and require the vendor to support the full workflow — records, QC, imaging, ultrasound — not just one reading task.
- If you research health-system AI adoption, county medical communities (县域医共体) are the most under-reported deployment surface in Chinese healthcare; Nanpi's conference materials are a starting paper trail.
- If you build medical AI for emerging markets, note the certification path Shukun used — NMPA Class III at home, then EU MDR and Japan PMDA — as a template for credibility-led export.