AI Diagnoses Entering Ordinary People: China's Inclusive Healthcare Service Upgrades

Securities Times Reporter Zhang Yifan

On April 9 this year, the State Council issued the “Several Measures to Accelerate the Construction of a Graded Diagnosis and Treatment System,” proposing to further promote the sinking of medical and health services and enhance grassroots capabilities. How can grassroots medical workers shift from passive referrals to being “able to read, and dare to make decisions,” thereby improving the quality of medical services? Some medical and health examination institutions have turned their attention to artificial intelligence (AI) technology.

“Early detection, early diagnosis, and early treatment.” AI is turning the vision of inclusive healthcare into everyday reality. With innovation in service scenarios, collaboration among industry, universities, and research institutes, and data enablement, China’s medical services are continuously extending to the grassroots and moving prevention forward—and it has also become a source of innovation for AI applications and the main frontline for practical deployment.

Moving the Front Gate of Medical Services Forward

In Changsha, 67-year-old diabetes patient Wang Huizhen (pseudonym) has personally experienced the convenience of “getting eye care at the doorstep.”

“Please look ahead—don’t blink.” At the Muyun Street Community Health Service Center in Tianxin District, Changsha, Wang Huizhen completed fundus photography under the guidance of a doctor. In less than two minutes, an analysis report automatically generated by AI was synchronized to her phone via a mini-program.

During this examination, AI alerted that Wang Huizhen showed symptoms of retinal bleeding. The community doctor immediately transferred her to Changsha Medical Center through the green channel of the medical alliance. After minimally invasive surgery, Wang Huizhen can return for follow-up at an outpatient department only 500 meters from her home, while the lead surgeon guides her medication remotely through video.

“The biggest value of this ‘AI-assisted fundus imaging diagnostic system’ is that it turns ‘early detection, early diagnosis, and early treatment’ into reality from a technical standpoint,” said Dai Weiwei, Executive Deputy Director of the Aier Digital Ophthalmology Research Institute, in an interview with reporters. In the past, patients without obvious symptoms would not see a doctor, and the specialized experience of community doctors at the grassroots level also varied. Now, the AI system can accurately screen 11 eye diseases within 2 minutes, and—across more than a thousand outlets in cooperation with Aier Eye Hospital (300015)—it has been called over 600,000 times in total.

To further promote the sinking of medical and health services and enhance grassroots service capacity, there are many medical institutions like Aier Eye Hospital that are taking AI technology as a key initiative.

For example, Meinian Health (002044) has introduced an ultrasound AI full-quantity quality control system in the examination process, automatically evaluating more than 2 million ultrasound images per day; AI imaging-assisted diagnosis covers over 90% of imaging examination projects, improving early screening efficiency by 40%. “AI applications effectively reduce the risk of missed diagnoses and ease bottlenecks in grassroots medical efficiency,” said a relevant person in charge at Meinian Health. “Taking AI-assisted small intestinal capsule endoscopy as an example, the AI system increased the detection sensitivity for a single lesion from 76.89%, based on traditional manual reading, to 99.90%.”

As the world’s only AI robot that has passed the national physician licensing exam written test, iFlytek Medical’s Zhiyi Assistant has, in services across more than 77,000 grassroots medical institutions, already provided more than 1.1 billion AI-assisted consultation recommendations and helped generate more than 450 million standardized electronic medical records.

Clinical Data Makes AI More Usable

The in-depth development of “AI + medical and health” applications has become a trend. China aims to achieve broad application by next year of intelligent assistance for grassroots diagnosis and treatment, intelligent assistance for clinical specialty and disease diagnosis decision-making, and intelligent patient-visit services in medical and health institutions. To this end, medical institutions, scientific research institutes, and hardware enterprises are working together, and offline medical service outlets that are most closely aligned with clinical needs and have the richest data accumulation are becoming “AI + medical and health” application innovation scenarios.

During the development of the aforementioned “AI-assisted fundus imaging diagnostic system,” Aier Eye Hospital formed a three-party cooperation model with the Institute of Computing Technology, Chinese Academy of Sciences, and medical device manufacturers, with clearly defined division of labor and complementary strengths—promoting the R&D of medical AI hardware and the transformation of results. In that cooperation, Aier Eye Hospital assumes the role of medical support. Teams of senior physicians complete tasks such as lesion annotation, disease classification, and clinical logic verification to ensure that the AI’s judgments comply with clinical diagnosis and treatment norms.

Dai Weiwei told Securities Times that clinical data and medical annotations are the key to ensuring that AI models “learn accurately.” To train high-quality fundus AI models, dozens of senior ophthalmologists participated in image annotation. Annotating a single image takes 5 to 10 minutes. In total, they completed tens of thousands of data annotations, providing high-quality support for algorithm training. Currently, Aier Eye Hospital’s physician team is also forming a special group targeting rare fundus disease groups, pushing AI into more subdivided and more complex areas.

“AI diagnostic models are not rare. The value of Aier Eye Hospital lies in our ‘1+8+N’ strategy: covering nearly a thousand medical outlets nationwide, building a plentiful professional talent team, and continuously accumulating from the annual scale of millions of ophthalmology outpatient visits and hundreds of thousands of ophthalmic surgeries,” Dai Weiwei said.

“Artificial intelligence technology needs to return to the essence of healthcare and serve clinical practice,” emphasized a relevant person in charge at iFlytek Medical. Currently, based on the Zhiyi Assistant—which has served more than 77,000 grassroots medical institutions and more than 600 hospitals through smart hospital solution services—iFlytek Medical can receive more than 1.7 million real diagnosis and treatment data entries returned every day. In 2025, the total volume of high-quality training data will increase by 30%, the medical knowledge base is approaching 1 billion articles, and this provides support for iteration of large-scale disease-specific models.

Turning the Service Flywheel

The prerequisite for the accelerated development of “AI + medical and health” applications is that data factors must be compliant, efficient, secure, and interconnected in an orderly manner. Only by, under strict compliance and privacy protection, realizing cross-facility and cross-device data integration, de-identification sharing, and co-development, can we support multi-center real-world research, disease risk model training, operational optimization, and supply chain collaboration.

As the developer of large models, a relevant person from iFlytek Medical said that the company currently uses domestically controlled computing power to achieve the end-to-end domestication of Xinghuo Medical large-model training and inference deployments, safeguarding medical data security from the root technical layer. With a full-stack autonomous and controllable technical system, it guards the bottom line of medical data security.

As an important supplier of medical data, Aier Eye Hospital is focusing on cleaning, governance, modeling, and analysis of massive data (603138), aiming to achieve the aggregation of high-quality data resources and gradually promote the assetization of data and the extraction of value. In 2025, Aier Eye Hospital has already integrated 66 disease-specific datasets focusing on specific eye diseases, including more than 300,000 cases of rare and key eye diseases, and completed the listing of 7 data products on data exchanges.

Currently, Dai Weiwei’s team at Aier Eye Hospital is working to tackle the technical difficulties in identifying rare diseases related to the fundus; the team at iFlytek Medical is striving to promote the deep integration of the Xinghuo Medical large model, the medical system, and intelligent hardware sensing terminals, attempting to establish a full-cycle health management service. With the joint efforts of many medical institutions, China’s healthcare services industry will continue to upgrade toward more intelligent, more systematic, and more accessible directions.

(Editor-in-charge: Dong Pingping)

     【Disclaimer】This article only represents the author’s personal views and is not related to Hexun. The Hexun website remains neutral with respect to the statements and viewpoints in the article, and does not provide any express or implied guarantees regarding the accuracy, reliability, or completeness of the content contained herein. Please note that readers should use this information as reference only and bear all responsibility themselves. Email: news_center@staff.hexun.com