{"slug": "china-just-let-ai-run-an-eye-clinic-here-s-what-happened", "title": "China Just Let AI Run an Eye Clinic. Here's What Happened.", "summary": "A team led by the Beijing Visual Science and Translational Eye Research Institute (BERI) in China ran an AI-agent-augmented eye clinic, AI-TEC, and reported in Nature Medicine that the AI's diagnostic accuracy on eye scans exceeded an AUROC of 0.93 after ophthalmologists supplied 1,426 high-quality labeled images, beating the original training set of nearly 27,000 lower-quality images. Clinical use of the AI-TEC processes fell to 41 of 1,113 examinations (3.8 percent) in a snapshot taken after five months, then climbed to 259 of 1,126 examinations (23 percent) the next month after the researchers reduced clicks and manual input. The researchers concluded that AI-native healthcare is \"fundamentally an ecosystem challenge\" depending on data quality, workflow interaction, clinician engagement, governance and monitoring, and measurable clinical value rather than algorithmic performance alone.", "body_md": "It's difficult to avoid the [discussions around AI](https://www.sciencealert.com/artificial-intelligence) at the moment – and how useful or otherwise it can be.\n\nScience and healthcare are two of the areas where the tech promises to be transformative.\n\nWe're already seeing AI used extensively in terms of [analyzing scans](https://www.sciencealert.com/ai-can-spot-pancreatic-cancer-years-before-diagnosis-study-finds), [designing drugs](https://www.sciencealert.com/ai-designed-drug-makes-patients-blood-look-biologically-younger-study-shows), and [predicting the future](https://www.sciencealert.com/ai-experiment-generated-40-000-hypothetical-bioweapons-in-6-hours-scientists-warn), but researchers led by a team from the Beijing Visual Science and Translational Eye Research Institute (BERI) in China wanted to take the deployment of AI further – by letting it run an eye clinic (still with oversight).\n\nThe clinic was called the AI-TEC (AI-Agent Augmented Tsinghua Eye Clinic), and while human doctors were still involved, it was designed from the ground up to maximize AI use rather than having these tools bolted on top of existing systems and practices.\n\nAs reported in [*Nature Medicine*](https://doi.org/10.1038/s41591-026-04631-z), that meant AI was used in everything from pre-consultation to patient follow-up, including the all-important eye scans themselves.\n\n\"We provide early implementation lessons in moving AI beyond algorithm performance toward clinical workflow transformation and system-level integration, where clinical value is ultimately created,\" [write](https://doi.org/10.1038/s41591-026-04631-z) the researchers in their published paper.\n\nSeveral interesting findings came out of the AI-TEC trial. Firstly, the AI started off with a relatively low success rate at identifying diseases in eye scans – diseases like glaucoma and age-related [macular degeneration](https://www.sciencealert.com/revolutionary-prosthetic-eye-chip-restores-sight-in-medical-first).\n\nHowever, the accuracy levels increased significantly when expert ophthalmologists fed the AI with 1,426 high-quality eye scan images correctly labeled with the relevant conditions.\n\nThis fresh data worked better than the original training data of almost 27,000 images that were of lesser quality and less comprehensively labeled.\n\nUsing a standardized diagnostic metric called AUROC, the AI was able to reach an accuracy of over 0.93 using the updated training data – roughly [on a par](https://doi.org/10.1038/s41586-023-06555-x) with other state-of-the-art scanning systems.\n\nSomething else the researchers found was that clinical usage of the AI tools started off high and then dipped. In a snapshot taken after five months, only 41 of 1,113 examinations (3.8 percent) in the month used the AI-TEC processes available.\n\nThis climbed back to 259 of 1,126 examinations (23 percent) the following month, after the researchers made the system faster and easier to use, with fewer clicks and less manual input required from staff.\n\n\"Our early experience showed that implementing an AI-native healthcare is fundamentally an ecosystem challenge,\" [write](https://doi.org/10.1038/s41591-026-04631-z) the researchers.\n\n\"Its effectiveness did not depend on any individual algorithmic performance but instead depended on data quality, workflow interaction, clinician engagement, adequate governance and monitoring, and measurable clinical value.\"\n\nAnother observation from the researchers: regular feedback from clinicians needs to happen quickly in order for the AI to improve, rather than it being supplied weeks later.\n\nThose are the three big takeaways: the need for quality data, the need for optimized operation, and the need for regular, rapid feedback.\n\n\"We emphasize the need for close on-site engagement and collaboration between clinicians and AI researchers,\" [write](https://doi.org/10.1038/s41591-026-04631-z) the researchers.\n\nIt's early days for this kind of AI-first clinic, but the results of this small-scale trial show they can be successful when the right ingredients are combined.\n\nPlenty of challenges remain, however, including resolving the way that AI focuses on results first (does patient A have an eye disease or not), whereas doctors focus on symptoms first (patient A reports blurry vision).\n\nIn other words, having an AI that's accurate at analyzing scan images doesn't necessarily translate into something that improves patient care. In fact, future tests of this kind may be better off not focusing on raw performance scores alone, the study suggests.\n\n**Related: [Scientists Find a Subtle Clue to ADHD Hidden in The Eyes](https://www.sciencealert.com/scientists-find-a-subtle-clue-to-adhd-hidden-in-the-eyes)**\n\n\"The success of 'AI-native' healthcare system should ultimately be measured by whether clinical workflow is transformed and whether AI improves health outcomes, rather than by whether AI achieves superior benchmark performance,\" [write](https://doi.org/10.1038/s41591-026-04631-z) the researchers.\n\nThe research has been published in [*Nature Medicine*](https://doi.org/10.1038/s41591-026-04631-z).\n\nThis article was fact-checked by [Fiona MacDonald](https://www.sciencealert.com/fiona-macdonald) and edited by [Fiona MacDonald](https://www.sciencealert.com/fiona-macdonald). While we pride ourselves on our process, we are only human. If you spot a mistake, [please let us know](https://www.sciencealert.com/contact-us).", "url": "https://wpnews.pro/news/china-just-let-ai-run-an-eye-clinic-here-s-what-happened", "canonical_source": "https://www.sciencealert.com/china-just-let-ai-run-an-eye-clinic-heres-what-happened", "published_at": "2026-09-14 10:00:00+00:00", "updated_at": "2026-09-14 10:05:06.065353+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-agents", "ai-research", "ai-products"], "entities": ["Beijing Visual Science and Translational Eye Research Institute", "AI-TEC", "Nature Medicine", "Tsinghua Eye Clinic"], "alternates": {"html": "https://wpnews.pro/news/china-just-let-ai-run-an-eye-clinic-here-s-what-happened", "markdown": "https://wpnews.pro/news/china-just-let-ai-run-an-eye-clinic-here-s-what-happened.md", "text": "https://wpnews.pro/news/china-just-let-ai-run-an-eye-clinic-here-s-what-happened.txt", "jsonld": "https://wpnews.pro/news/china-just-let-ai-run-an-eye-clinic-here-s-what-happened.jsonld"}}