Editor's note, have you AI today? Nowadays, it seems that our lives and work are increasingly inseparable from digital technologies such as artificial intelligence (AI), and technology is still "growing" every day. From "digital intelligence+consumption", "digital intelligence+cultural tourism" to "digital intelligence+sports", from autonomous driving, intelligent manufacturing to smart cities, digital intelligence technology is accelerating its integration into various industries, constantly opening up new application scenarios and continuously changing the way humans produce and live. Starting from today, this edition will open the "Decoding Intelligence+" column, allowing readers to explore the constantly emerging new scenarios of digital technology applications and observe the infinite possibilities they bring. From guidance robots to tumor early screening in medical imaging, the application scenarios of artificial intelligence (AI) technology in the medical field are becoming increasingly diverse. In the wave of artificial intelligence big models, "asking AI when sick" has attracted particular attention. When patients enter the consultation room with AI generated diagnosis and treatment suggestions, and when AI's "opinions" even challenge doctors' judgments, a series of questions urgently need to be answered: Is AI healthcare reliable? Will it replace doctors? How can patients and doctors avoid risks while embracing efficiency? AI becomes a good helper for diagnosis and treatment. Open the "AI Hepatobiliary Hospital" in the WeChat mini program of Beijing Tsinghua Chang Gung Hospital, and the reporter enters "discomfort in the upper right abdomen" in the dialog box. Soon, the AI started chatting with the reporter: "Are your symptoms persistent or intermittent?" "Are they accompanied by fever or nausea and vomiting?" After a few questions, the AI provided recommendations for diagnosis and treatment in the hepatobiliary department. Tsinghua Changgeng Hospital is developing a comprehensive liver disease management model, and the preliminary version has been launched on the hospital's WeChat mini program. Currently, it can conduct pre diagnosis based on patients' symptoms and provide triage recommendations. ”Yang Ming, Chief Physician of Hepatobiliary Department at Tsinghua University Beijing Tsinghua Chang Gung Hospital, introduced that this AI system will provide triage recommendations based on patients' symptoms and laboratory tests, with high accuracy. During the interview, many doctors reported that there are increasingly more cases of patients coming to seek treatment with AI diagnosis and treatment recommendations. Some patients will use AI to sort out their thoughts before seeking medical treatment, with a relatively clear approach. ”Chen Xiuyuan, Deputy Chief Physician of the Department of Thoracic Surgery at Peking University People's Hospital, told reporters that patients will use AI big models to obtain preliminary disease explanations and possible diagnosis and treatment directions based on their medical history and examination data. This is equivalent to conducting a disease education for patients in advance, giving them a preliminary understanding of the disease, making it easier for them to understand the doctor's professional judgment and advice, and making subsequent communication smoother and more efficient. ”Yang Ming said. Patients have AI for consultation, and doctors also use AI for medical treatment. Several doctors have stated in interviews that for nodules with a diameter less than 5 millimeters, doctors have a lower detection rate, but the detection rate has significantly improved after using AI. Very useful, "said Wang Yi, chief physician of the Radiology Department at Peking University People's Hospital. If surgery is compared to driving, then CT is like an accurate paper map, and with AI 3D reconstruction, it is equivalent to having a more precise and intuitive electronic map. ”The "AI 3D reconstruction" algorithm mentioned by Chen Xiuyuan has been deployed at Peking University People's Hospital for many years. This system uses AI to present complex lung structures more accurately, and the accuracy of anatomical structure recognition has been improved. The application of such systems can help doctors free themselves from the heavy workload of initial imaging screening, and focus more energy on conducting more in-depth comprehensive analysis based on imaging results, developing personalized treatment plans, and dealing with more complex diagnostic problems. ”Zhang Yunzeng, Deputy Chief Physician of the Second Department of Thoracic Surgery at Shandong Provincial Public Health Clinical Center, said. Not only in medical imaging, but also in surgical planning, AI has unexpected effects. Li Haifeng, Deputy Chief Physician of Joint Surgery at the Orthopedic Department of the General Hospital of the Chinese People's Liberation Army, introduced joint replacement surgery as an example: "In the past, such surgeries often required preparing a complete set of different types of prostheses for each patient to prevent model mismatch, resulting in resource waste. Now, AI can accurately predict the required prosthesis model in advance by analyzing patients' CT data and combining it with massive past surgical data models. ”Information overload can easily exacerbate anxiety. AI has shown certain efficiency and accuracy in the process of assisting medical care. Will it replace doctors? During the interview, although doctors generally recognized the value of AI in assisting diagnosis and treatment, they still held a cautious attitude towards the specific conclusions or treatment recommendations provided by AI. I only recommend patients to use AI consultation as a way to understand the disease, and do not recommend patients to follow AI's advice. ”Li Xiaohong, chief physician of the Department of Spleen, Stomach, Liver and Gallbladder at Dongfang Hospital of Beijing University of Traditional Chinese Medicine, said. Regarding this, Yang Ming explained: "Currently, AI diagnosis is mainly based on large models, and the captured data will have a significant impact on its generated results. ”These contents are indeed very logical and systematic, but whether they are applicable to different patients still needs further judgment. ”Wang Yi pointed out that if patients do not have much knowledge about the disease, it may be difficult to identify the problem. Regarding medical advice for AI, many experts have stated that "overloaded" information can actually exacerbate patients' anxiety. Li Haifeng said, "Sometimes patients use AI generated reports that are very detailed, listing all possible problems, causing them to seek confirmation from doctors with confusion or even panic. However, the reality is that many assumptions do not have clinical significance. ”Some diseases are systemic problems caused by multiple factors, and it is difficult to make an accurate diagnosis solely based on the symptom description provided by the patient. ”Li Xiaohong frankly stated that as people's understanding of AI deepens, they gradually realize that the content generated by AI needs to be considered for authenticity. A prescription not only means prescribing the right medicine, but also carries the responsibility and commitment of a doctor behind it. Doctors use professional knowledge as the starting point for diagnosis and treatment, but when facing each patient, they need to adjust the treatment plan based on their own characteristics, and AI may be lacking in this regard. ”Yang Ming explained with liver disease as an example, for example, a patient with elevated transaminase had a history of hepatitis B and fatty liver, and recently had a history of heavy drinking and taking statins. "AI may only take the initiative to ask about their past medical history during diagnosis and treatment. Based on their information about hepatitis B, AI will give advice on 'need to take antiviral drugs', but it will miss key details that the patient did not take the initiative to provide, such as drinking and drug history, so the advice given is easily biased. ”In the field of imaging, although AI has been widely used, relying solely on AI may miss some very early and atypical lesions under current screening technology conditions. ”Zhang Yun once stated that without comprehensive judgment based on multidimensional information such as medication history and past imaging comparison, misdiagnosis is likely to occur. Big models can free hands and feet, but they cannot replace the brain. ”Li Xiaohong believes that medical decision-making relies on complex clinical judgments and rich experience, especially in the face of atypical cases or coexistence of multiple diseases. Experienced doctors can capture subtle symptoms and signs, which is currently difficult for AI to achieve. Doctors are not only healers of diseases, but also psychological supporters of patients. ”Zhang Yun once bluntly stated that "AI is difficult to provide psychological support for patients, while medicine has a temperature that is transmitted through doctors. ”To make AI more 'compliant' and 'transparent', we need to be wary of the phenomenon of 'AI illusion', which refers to the fabrication of information, when AI models generate content. In clinical use, AI occasionally makes some ridiculous mistakes. ”Zhang Daoqiang, Dean of the School of Artificial Intelligence at Nanjing University of Aeronautics and Astronautics, gave an example that the changes in clinical imaging are extremely complex. Sometimes, what is seen may be "interference signals" without diagnostic value, but AI may judge them as lesions. "Some users also find that there are situations where AI assists in generating content, such as fabricating medical terminology sources and fictitious references. ”Regarding the issue of information fabrication, Zhang Daoqiang believes that improvements should be made from both algorithm and data aspects. The medical field is very special and requires strict control of errors. This requires us to strictly control the uniqueness and accuracy of data in the early stage of large model development. Taking algorithms as an example, improving the anti-interference and reliability of AI is an important issue. How to improve the recognition accuracy and reliability of the system in real environments when AI enters from the laboratory? Any change in information may cause subtle deviations. ”Zhang Daoqiang said. Interpretability is also the development direction of AI healthcare. Professor Qin Jie from the School of Artificial Intelligence at Nanjing University of Aeronautics and Astronautics explained: "The decision-making process of AI is more like a 'black box', and patients may not be able to determine how the results were obtained. Therefore, the path of decision-making should be explained to help people make better judgments. Our thinking direction is to make AI itself more 'rule-based' and 'transparent'. ”For data samples, the size and quality of the samples will have a significant impact on the results generated by AI. How to better integrate the experience of doctors with data-driven big models? How to replicate the experience of top doctors on AI? These are all things we need to study. ”Qin Jie said that in terms of the tasks and performance of the model, "combining general and specialized knowledge" is the next development direction of AI, "that is, to deeply explore tasks based on large models and vertical scenarios." In the face of AI, we should actively embrace new technologies while maintaining rationality. AI is just a tool, definitely not a lazy tool. ”Wang Yi stated that when doctors use AI, they should combine their solid professional foundation and rich clinical practice to critically consider and rationally analyze the information provided by AI, and not overly rely on it. During the interview, experts also suggested that relevant departments should integrate medical big data, conduct unified research and design, establish standards and evaluation systems for AI doctors, and enable AI to better assist doctors. From a legal perspective, it is more difficult to determine and attribute responsibility for AI related cases. ”Ma Yide, professor of the School of Public Policy and Management of the University of Chinese Academy of Sciences, pointed out that "the development and deployment of AI applications often involve multiple links and subjects. From algorithm design to specific operations, every step may have an impact on the final AI behavior. The elongation and dispersion of the responsibility chain make it difficult to identify the responsible parties when problems arise. ”Ma Yide suggests promoting the standardized deployment of AI medical applications through measures such as improving laws and regulations, strengthening data security, establishing accountability mechanisms, and enhancing ethical supervision. At present, the National Medical Products Administration has clarified that medical software that uses AI for disease diagnosis, decision-making assistance, image recognition, etc. belongs to the category of medical devices and must be registered and regulated in accordance with medical device regulations. Patients use AI for consultations, and AI answers do not require legal responsibility, while doctors are responsible for the patient's diagnosis and treatment results. ”Yang Ming reminds us to ensure that AI technology is applied reasonably within the legal framework, strictly adhere to data security and ethical bottom lines, and ensure that technology applications always serve the essence of healthcare. (New Society)