AI rehabilitation therapy robot at the 16th Hong Kong International Medical and Healthcare Exhibition. How will artificial intelligence change the healthcare service scene? The Implementation Opinions on Promoting and Standardizing the Development of "Artificial Intelligence+Medical and Health" Applications (hereinafter referred to as the "Implementation Opinions") recently released by the National Health Commission and five other departments have drawn a development "roadmap": by 2027, intelligent assistance for grassroots diagnosis and treatment, intelligent assistance for clinical specialty diagnosis and treatment decision-making, and intelligent services for patient visits will be widely applied in medical and health institutions; By 2030, the application of intelligent assistance in grassroots diagnosis and treatment will basically achieve full coverage. According to this "roadmap", artificial intelligence will continuously enrich its application scenarios in the medical and health field, enhance service capabilities, ensure service safety, optimize resource allocation, innovate continuous intelligent services throughout the entire chain of prevention, diagnosis and treatment, rehabilitation, and health management, and better meet the growing health service needs of the people. The Implementation Opinions mark a new stage in which artificial intelligence healthcare has moved from pilot exploration to large-scale promotion. ”Li Nan, the director of Handan People's Hospital in Hebei Province, said, "We hope to ultimately build an intelligent medical system that improves efficiency with AI, brings doctors together, and benefits patients. ”Focusing on grassroots applications, the State Council issued the "Opinions on Deepening the Implementation of the 'Artificial Intelligence+' Action 'in August this year, which proposed exploring and promoting high-level resident health assistants that everyone can enjoy, orderly promoting the application of artificial intelligence in auxiliary diagnosis and treatment, health management, medical insurance services and other scenarios, and significantly improving the capacity and efficiency of grassroots medical and health services. On this basis, the "Implementation Opinions" have been issued, which clarify 24 key applications of artificial intelligence in 8 directions: grassroots application, clinical diagnosis and treatment, patient service, traditional Chinese medicine, public health, scientific research and teaching, industry governance, and health industry. Among them, "artificial intelligence+grassroots applications" is the primary direction. The Implementation Opinions propose to establish intelligent assisted diagnosis and treatment applications for grassroots doctors. Establish intelligent auxiliary diagnosis and treatment applications for common and frequently occurring diseases at the grassroots level, providing intelligent applications such as auxiliary diagnosis and treatment, prescription review, follow-up management, and traditional Chinese medicine diagnosis and treatment to grassroots doctors, and enhancing their service capabilities in general auxiliary diagnosis, disease differentiation diagnosis, and medical imaging auxiliary diagnosis. Currently, AI has a technological foundation at the grassroots level. AI is no longer just at the conceptual stage, but has actually entered various scenarios in hospitals. ”Li Nan gave an example that in the diagnostic process, AI can automatically analyze patient symptoms, generate preliminary diagnostic recommendations, and assist doctors in writing standardized medical records, significantly improving the standardization rate of medical records; In the treatment process, AI can provide reference for treatment plans. In some areas, by building "electrocardiogram networks" and "imaging networks", rural residents can enjoy expert level diagnosis and treatment plans from tertiary hospitals locally; In complex scenarios such as hospitalization, some advanced medical models can even reduce doctors' medical record writing time by half. The model of artificial intelligence participating in "chain style" medical and health services has begun to take shape. Previously, I had to spend a lot of time responding to basic consultations every day, but now I have an intelligent agent 'avatar' that can answer many routine questions from patients at any time. ”Lin Zhimiao, Deputy Director of the Institute of Dermatology and Venereology at Southern Medical University, shared the changes brought about by expert doctors' intelligent agents. He introduced that relying on JD Health's "Jingyi Qianxun" medical model, the intelligent agent can not only accurately respond to common problems of psoriasis and rare skin diseases, but also proactively book his outpatient time for patients who need further diagnosis and treatment. This frees doctors from repetitive consultations, allowing them to receive 3-5 more difficult cases every day, and improves diagnosis and treatment efficiency by at least half. Li Nan added that artificial intelligence can also help doctors improve their professional abilities. Artificial intelligence can provide real-time reference opinions, such as alerting doctors of suspicious lesions when reading films and reminding them of medication contraindications when prescribing medication. This is equivalent to continuous "online training" and can continuously "recharge" grassroots doctors. Highlighting scenario driven approach, the Implementation Opinions adhere to scenario driven approach, targeting real business in the health industry, relying on real scenarios, and addressing real needs. It is worth noting that the Implementation Opinions propose to establish a batch of high-quality datasets and trustworthy data spaces for the health industry by 2027. In Li Nan's view, to achieve this goal, the hospital information department, medical record room, and clinical department need to collaborate and clean, desensitize, and structure historical medical records, imaging data, pathological reports, and genomic data according to unified standards, forming a high-quality specialized disease dataset, and establishing a secure and controllable data exchange and sharing platform within the hospital and among medical consortia. The relevant person in charge of JD Health Exploration Research Institute said that high-quality datasets need to balance the professionalism of medical specialties and dataset inclusion. To this end, JD Health has collaborated with several top medical institutions in China to build training data for specialized and disease specific models based on a large amount of high-quality clinical real data from multiple dimensions such as medical history, laboratory tests, imaging, and pathology, and match them with real medical scene evaluations. Based on professional databases, AI big model developers are also continuously developing specialized disease big models covering common tumors and major chronic diseases, helping areas with scarce medical resources to enjoy high-level medical expert services. The working mode of hospitals at all levels will undergo a profound transformation from 'experience driven, labor-intensive' to 'data-driven, human-machine collaborative'. ”Li Nan stated that the "Implementation Opinions" provide guidance for hospitals to promote intelligent transformation at multiple levels. In the future, top hospitals will focus more on solving difficult and complicated diseases, conducting cutting-edge scientific research, and developing clinical guidelines. Routine diagnosis and treatment, chronic disease management, and postoperative follow-up will mainly be completed by grassroots hospitals and family doctors assisted by artificial intelligence. Standardizing safety supervision and ensuring safety has always been the lifeline of the healthcare industry. The "Implementation Opinions" will separate a section on "regulating security supervision", requiring measures such as optimizing industry management and audit systems, innovating regulatory methods and warning mechanisms, strengthening data security and personal privacy protection to ensure the safety, reliability, and controllability of artificial intelligence in the medical field. The security risks of artificial intelligence in the medical field exhibit diverse and intertwined characteristics, including inherent challenges such as algorithmic black boxes and data deception, as well as new security risks arising from the development trends of multimodal collaboration and intelligent agent interconnection. ”Gong Mengchun, director of the Multi Modal Data Fusion Application Laboratory (GMCLab) at Guangdong Medical University, told reporters that the improvement of risk control mechanisms requires the "three pillars" of technological innovation, ethical review, and artificial intelligence literacy enhancement. The Implementation Opinions require the improvement of a comprehensive governance mechanism that includes government supervision, institutional autonomy, industry self-discipline, and social supervision. This has already made progress in some medical institutions. Gong Mengchun introduced that the Expert Consensus on Clinical Ethical Governance of Generative Medical AI (GMAI) (2025) jointly prepared by experts from the Chinese Academy of Medical Sciences, Guangdong Medical University, West China Hospital of Sichuan University, the Institute of Automation of the Chinese Academy of Sciences, Shenzhou Medical Technology Co., Ltd. and other units clearly proposed to implement mandatory confidence scoring and dynamic hallucination threshold control on artificial intelligence systems. This echoes the requirements of "conducting application monitoring and evaluation" and "establishing large-scale model application evaluation and verification" in the Implementation Opinions. In response to public concerns about data security and privacy protection, the Implementation Opinions also explicitly require the establishment and improvement of an intelligent application data security protection system to promote standardized circulation and sharing of data. Gong Mengchun said that federated learning is a technology that promotes "win-win cooperation" while also "protecting privacy". Through this technology, data is locked in separate "safes" in different hospitals, and only AI models, as "students," flow and learn between hospitals. Based on this' data doesn't move, model moves' model, multiple hospitals can jointly train a powerful AI model without sharing any raw data, ensuring data security and privacy protection. (New Society)