Which "AI Doctors" Are Already on Duty? (Big Health Observation) Eye diseases, CT scans, chronic disease management... Today, a batch of specialized intelligent agents developed by top-tier tertiary hospitals have stepped out of laboratories and are being deployed for daily clinical services in community health centers, county-level hospitals, and even remote high-altitude regions. Recently, these "AI doctors" referred to by the public took center stage at the phased results launch event of the National Medical Field Artificial Intelligence Application Pilot Base. The event prioritized hospitals, doctors, and enterprise users, showcasing the warmth and innovative value of deep integration between artificial intelligence and healthcare from the perspectives of AI users and beneficiaries. The story began a year ago. In May 2025, Beijing launched the construction of the first batch of medical field national artificial intelligence application pilot bases. Under the overall guidance of the National Development and Reform Commission and the National Health Commission, the project advanced through a mechanism of department-city linkage, city-district coordination, and government-enterprise cooperation. Currently, the base has built a common support capability system covering the "five-in-one" framework of computation, data, models, applications, and security: it has established a 384P domestic AI computing resource pool, constructed a trustworthy healthcare data space, set up an application product marketplace aggregating over 100 intelligent agents, and created an integrated security system for "cloud, network, data, models, and applications," fully connecting the entire closed-loop chain from technology R&D, clinical validation, industrial transformation, to scenario implementation. Building a Solid Security Bottom Line to Activate the Smart Brain Balancing security and efficiency is the core prerequisite for the practical application of medical artificial intelligence. In chronic disease prevention and control, Yang Li, Vice Dean of Peking University First Hospital, introduced that the hospital's developed chronic kidney disease comorbidity digital doctor assistant adopts a "dual-engine" architecture: The rule engine is responsible for "avoiding errors," serving as the system's security bottom line—ensuring patient safety through crisis value warnings, key indicator identification, and interception of unreasonable prescriptions. The large model engine is responsible for "understanding people better," acting as the multi-agent smart brain—simulating multidisciplinary expert consultations for kidney disease, hypertension, diabetes, etc., to provide individualized recommendations. This intelligent agent has been deployed in community institutions in Xicheng District, Beijing, covering over 1 million people across 3 provinces. After application, community doctors achieved a 100% mastery rate in kidney disease screening and diagnosis.

Addressing Radiology Report Backlogs and High Missed Diagnosis Risks Liu Yaeou, Director of the Radiology Department at Beijing Tiantan Hospital, introduced "Xiao Jun Doctor 2.0"—a full-process diagnostic system covering nearly 100 diseases in head CT scans. It shortens report generation time to under 1 minute, achieves a diagnostic accuracy rate of nearly 80%, and can identify some complex cases at a level surpassing even resident doctors or senior physicians. Currently applied in over 40 medical institutions across 14 provinces and cities nationwide, it has formed a new workflow of "AI drafting initial reports, doctors reviewing and finalizing them." High-Quality Medical Resources Breaking Through Mountains and Seas From bustling metropolises to snow-covered plateaus, artificial intelligence is breaking spatial and temporal barriers, making high-quality medical services more equitable and accessible. "Doctors who can diagnose rare diseases are rarer than the patients themselves." said Zhou Xiang, Director of the Information Center at Peking Union Medical College Hospital (PUMCH). The hospital’s specialized intelligent agent matrix covers over 110 agents in fields such as rare diseases, cardiopulmonary imaging, tumor radiotherapy, and gynecological pelvic floor disorders, and has been deployed to county-level and township medical institutions across more than 30 provinces nationwide. Zhou Xiang added that the PUMCH Taichu Rare Disease Auxiliary Diagnosis and Treatment Full-Process Intelligent Agent, which integrates patient-side, doctor-side, clinical testing, and gene testing capabilities, achieves 99% accuracy in genetic variation analysis and improves collaboration efficiency by 33% compared to manual methods. It has already been promoted to 419 hospitals. Beijing Anzhen Hospital’s cardiovascular disease ultrasound diagnosis and decision-making intelligent agent has built a full-process ultrasound image perception and disease reasoning AI model. It has been deployed in over 1,000 medical institutions across 31 provinces, regions, and municipalities, including remote western areas like Tibet, Xinjiang, Ningxia, and Qinghai, benefiting 1.3 million newborns. Liang Jianbiao, Director of the Ultrasound Department at Lhasa People’s Hospital, said: "With Anzhen’s AI large model, Lhasa now has the 'brain' of Beijing experts. These large models have seen far more cases than a doctor could encounter in their entire lifetime."

"Flowers Bloom Inside the Wall, Fragrance Spreads Beyond" Robust technical capabilities not only support the enhancement of grassroots diagnostic capabilities domestically but also serve as a hallmark of "China-made" going global. "We've achieved 'flowers blooming inside the wall, fragrance spreading beyond,'" said Yuan Jin, President of Beijing Tongren Hospital. "Intelligent agents for screening blinding eye diseases and major chronic conditions have been deployed in Southeast Asia, Gulf countries, and other regions. Recently, we even received procurement requests from a South American country—this is a practical demonstration of Chinese technology and China-made products moving overseas." This system, capable of "one-eye dual screening" (detecting blinding eye diseases while warning of systemic chronic diseases like hypertension and diabetes), has been implemented in 30 provinces and 140 prefecture-level cities, serving over 30 million people cumulatively. "In the past, it took an average of 5 to 8 years for ophthalmic AI products to go from paper publications to market launch—the missing link was a standardized pilot testing environment," Yuan Jin explained. Relying on the pilot base, the "Three Standards and Two Reviews" quality control system has enabled multiple products to complete pilot verification and enter the market, reducing R&D costs by 40% and shortening the time-to-market. Another product heading international is the AI pediatrician intelligent agent from the Children's Hospital of Capital Medical University. Zhao Chengsong, Executive Vice President of the hospital, introduced that to address insufficient grassroots pediatric resources, the grassroots version of the AI pediatrician intelligent agent has been deployed in 8 provinces, leaving grassroots medical institutions with a "medical assistant that can't be taken away"—capable of assisting in the diagnosis and treatment of common and prevalent diseases, indicating referral criteria, and standardizing clinical practices. In February this year, the all-in-one machine equipped with this intelligent agent won the International Red Dot Design Award. Multi-language versions (English, Russian, French, etc.) have been developed, with the Russian version officially launched at the Russia-China Expo this year.

Biomedical Manufacturing Equipped with an Accelerator Beyond clinical diagnosis and treatment, artificial intelligence is playing a pivotal role in the biomedical manufacturing sector. From drug discovery to preclinical research, from clinical trials to new drug launches, multiple self-developed AI pharmaceutical models have installed accelerators for the entire drug R&D chain, helping shift the industry from "finding needles in haystacks" to "following a map to find targets." The OpenComplex Plus unified drug development model released by the Beijing Institute for Artificial Intelligence (BAAI) connects four key links: target discovery, structure prediction, affinity assessment, and high-throughput reverse screening, achieving performance at the international leading level. China Telecom, leveraging the pilot base, has built an AI+ Drug R&D Public Service Platform, which has connected 33 drug pipelines from 17 domestic and international pharmaceutical companies. Among them, the accuracy of intelligent patient recruitment has improved from 25% (using traditional methods) to 85%. Wang Wei, Vice President of Bayer’s Prescription Drug Division, stated: "The 'policy + platform + data + scenario' full-chain support system built in Beijing makes medical big data 'accessible, controllable, verifiable, and actionable,' providing core support for enterprise innovation." At the conference, Zhang Jianrong, General Manager of China Unicom Jingyi Large Model Company, announced the launch of the "High-Quality Datasets 100 Million Yuan Support Plan." A special fund exceeding 100 million yuan has been established to solicit medical data from six categories—including clinical diagnosis and medical imaging—from medical institutions, research institutes, and AI enterprises, aiming to break the bottleneck of medical AI data supply at the source. As the wave of intelligence surges and medical transformation reaches a critical moment, Guo Yanhong, Deputy Director of the National Health Commission, stated at the conference: "Building pilot bases is an important platform for implementing the national AI strategy and serving the layout of new quality productive forces in the health sector. In the next step, we will refine policies, strengthen coordination, promote differentiated development and result-sharing among various bases, avoid low-level repetitive R&D, and truly enable AI to 'conduct high-end research at the top and apply at the grassroots level,' benefiting the general public."