In recent years, "how AI (artificial intelligence) can truly empower healthcare" has become a high-frequency topic in the industry. Economic Information Daily reporter recently found that AI applications in specific scenarios such as medical record quality control, intelligent triage and clinical trial matching have been quietly implemented. Currently, AI has shown tremendous value in reducing the burden on doctors, optimizing patient experience, and promoting scientific research innovation, and has entered a critical period of "large-scale implementation" from "experimental exploration". How can AI truly enter hospitals? Several experts from fields such as hospital management, clinical frontline, and policy research shared their observations and practices. Accelerating the implementation of scenarios helps improve the quality and efficiency of medical services. In the past, doctors had to spend a lot of time just organizing medical history and verifying logic when writing an inpatient medical record. Now, through AI voice collection and automatic summarization and citation of historical medical records, work efficiency has been greatly improved. ”Zhang Qian, the director of Beijing Shijitan Hospital, introduced that the hospital's AI applications have covered multiple fields such as patient services, clinical assistance, and medical management. Among them, scenarios such as medical record quality control, intelligent pre diagnosis, and interpretation of test reports have been promoted throughout the hospital. Essentially, it allows doctors to focus more on diagnosis and treatment itself. ”Zhang Qian stated that the AI medical record connotation quality control system can not only warn of time limit issues in medical record writing, but also accurately identify logical errors before and after, helping doctors and quality control personnel quickly identify defects; Intelligent pre consultation automatically pushes reminders to patients after registration, guiding them to improve their medical history information in advance. These data can be directly written back to the doctor's workstation, greatly reducing outpatient consultation time. Since its launch, the usage has exceeded 100000 times. The value of AI empowering healthcare is also reflected in the high difficulty field of tumor diagnosis and treatment. Song Yuqin, Vice President of Peking University Cancer Hospital, frankly stated that the core value of AI for hospitals lies in "improving quality and efficiency", which not only reduces the clinical burden of doctors, but also provides strong support for scientific research and teaching. It is reported that in response to the difficulty of triage for cancer patients, the Lymphoma Department of Peking University Cancer Hospital has built an online AI consultation platform, which has served thousands of people since its launch, effectively solving the pain points of patients' "hanging up the wrong number and running the wrong route"; In the field of scientific research, AI assists in building specialized databases that can generate Chinese and English literature reviews with just one click. It can also quickly generate clinical research plans based on research focus, sample size, and other information, reducing the originally weeks long work to just a few hours. From a technical perspective, the potential applications of AI go far beyond that. ”Zhao Yu, a researcher at the Western Research Institute of Zhongke Computing Technology, stated that in the future, AI will fully penetrate all aspects of healthcare, covering scenarios such as smart medical records, imaging diagnosis, surgical robots, and autonomous drug delivery systems. He stated that AI can effectively expand human intelligence and physical abilities, for example, surgical robots can achieve sub millimeter level accuracy, and remote surgery can also use AI to reduce operational risks. Despite the prominent advantages of AI in accelerating clinical applications from passive to active, many hospitals have been plagued by the problem of "unwilling to use" and "afraid not to use" during the implementation process. Zhang Qian admitted that in the early stages of the AI system's launch, some doctors had resistance, and the usage rate of the medical record AI quality control system was not high. The outpatient voice medical record generation function made departments with complex conditions willing to try it, while the scenario of simple follow-up appointments and medication prescriptions was rarely sought after. The core reason lies in insufficient guidance from management policies and insufficient motivation for doctors to use them; In addition, the initial model has insufficient adaptability, and AI without specialized tuning is prone to inaccurate medical record generation, which makes doctors hesitant to use it. ”Zhang Qian said. To this end, Beijing Shijitan Hospital has strengthened the requirements for medical record quality management, gradually turning AI quality control into a "must-have tool" for doctors. By establishing a feedback mechanism, active departments can summarize problems and synchronize with the IT department to optimize models. When doctors discover that AI can help them reduce errors and save time, they will naturally shift from passive acceptance to active use. ”Zhang Qian said. This transformation is also reflected in the practice of Peking University Cancer Hospital. Zhao Jun, chief physician of the Department of Chest Oncology at Peking University Cancer Hospital, stated that the promotion of AI needs to be combined with clinical needs and cannot be a one size fits all approach. Taking pre consultation as an example, the hospital promotes implementation through proposals from the staff congress, optimizes processes based on the characteristics of cancer patients, and encourages patients to actively participate in information confirmation. This not only ensures data accuracy but also enhances patients' medical experience. He said, 'AI is not' forced 'on doctors, but needs to truly address their pain points in order to achieve benign promotion.'. ”Song Yuqin revealed that Peking University Cancer Hospital plans to invest tens of millions of yuan to upgrade its AI system this year, aiming to optimize technology and provide training guidance to enable more doctors to feel the value of AI. We encourage doctors to actively try and provide feedback on problems, forming a virtuous cycle of 'use optimization reuse'. Breaking through the three major obstacles and promoting the large-scale implementation of medical AI, AI empowers life and health, which is both a challenge and an opportunity, as well as a responsibility. ”Liu Yuanli, a counselor of the State Council and a long-term professor at the School of Health Management and Policy of Peking Union Medical College, believes that in order to achieve large-scale application of medical AI, it is necessary to overcome the three major obstacles of data, evaluation, and implementation. Industry experts generally believe that data is the "fuel" for AI, but the supply of high-quality, standardized, and multimodal medical and health data is insufficient, and a secure, efficient, and trustworthy data sharing and circulation mechanism needs to be improved, becoming the primary bottleneck restricting the development of AI. Liu Yuanli said that as data holders, public hospitals generally have the problem of "not being able, afraid, and unwilling" to share data. 'Cannot' is because medical data is multimodal, highly complex, and highly specialized, and currently hospitals generally lack mature data governance and development capabilities; 'I dare not' is because health and medical data are highly sensitive, privacy protection and security responsibilities are under great pressure, and there are many concerns about sharing; 'Reluctance' is due to the lack of reasonable incentives and value return mechanisms for data contributions. Currently, while large models are rapidly iterating and demonstrating enormous value, problems such as algorithm bias and the risk of misdiagnosis and missed diagnosis have also emerged. Liu Yuanli stated that the more advanced the technology, the more regulation needs to keep up. To make AI truly trustworthy, usable, secure, and controllable, we should accelerate the construction of an authoritative evaluation mechanism and platform for artificial intelligence that covers the entire chain of research and development, approval, application, and supervision, and use unified scientific and authoritative standards to define the safe boundaries of technological innovation. No matter how good the technology is, it only has value when it is truly put into use. ”Zhang Qian believes that the scaling up of medical AI requires top-level planning and policy guidance, promoting the integration and sharing of medical, medical insurance, and pharmaceutical data, activating data elements, and ensuring the engine foundation for AI development; Within the hospital, overall planning is also necessary to coordinate and promote the deployment of computing power, model selection, and scenario promotion and application. (New Society)