On November 11, Shanghai Oriental Hospital officially released the large artificial intelligence medical model "med-go", which can effectively assist doctors in analyzing complex and rare cases, provide accurate and transparent clinical diagnosis and treatment suggestions, and empower artificial intelligence and medical depth. Fusion.
"Personal chief physician" for grassroots doctors
At present, "med-go" has been installed in the doctor workstation of Dongfang Hospital and connected to the hospital's his system; in addition, "med-go" has been installed in 15 community health service centers in Pudong New Area within the medical consortium of Dongfang Hospital, Jiangsu Sheyang County People's Hospital and its affiliated community health service centers, and Shanxi Xinzhou People's Hospital and its affiliated community health service centers have also installed "med-go". After the doctor fills in the patient's chief complaint, history of current illness, physical examination, etc., "med-go" can provide auxiliary suggestions for diagnosis, differential diagnosis, and next step treatment, which not only improves the quality and efficiency of diagnosis and treatment, but also reduces the occurrence of missed diagnosis and misdiagnosis. , which performs particularly well in difficult and rare cases.
"med-go" was jointly developed by the expert team of Dongfang Hospital and the Biomedical Artificial Intelligence Joint Laboratory jointly established by the Software Institute of the Chinese Academy of Sciences. At the press conference on November 11, Zhang Haitao, founder of “med-go” and director of the Department of Emergency and Critical Care Medicine of Dongfang Hospital, demonstrated a case on the spot: This was a real case of pediatrics in a famous tertiary hospital in Beijing. The child was hospitalized twice. , it took nearly a year to make the correct diagnosis. It was a very rare autoimmune disease. Zhang Haitao entered the children's medical records into "med-go", and in a few minutes he gave a diagnosis that took experts a year to make; when he entered the same medical records into the best large-scale model in the United States, the conclusions reached included the correct Diagnosis, but with a few more options. Zhang Haitao said: "It would take more than an hour to read this pile of medical records to a real chief physician!"
As people's understanding of the disease deepens, medical disciplines become more and more detailed. Even the same breast cancer must be divided into different types. It is difficult for a chief physician to master the knowledge of each subspecialty. However, this is not a problem for computers. A "med-go" is an "all-round" doctor. For a patient's symptoms, "med-go" can "think" and make judgments from all aspects of the patient's internal and external aspects, and provide advice. Doctors provide assistance.
“med-go” has consumed more than 6,000 textbooks
Why is "med-go" so powerful? Because the "material" fed to it is not only sufficient, but also closely follows international and domestic authoritative teaching materials. Zhang Haitao said: "Currently, it is based on 20 billion high-quality medical data. We have used more than 6,000 textbooks to train it. Some of the internationally authoritative textbooks do not yet have Chinese versions. Several of our societies and more than 60 experts just Translated page by page, I dug out a new book with more than 10,000 pages and fed it to it. "Theoretically, "med-go" has newer and more complete knowledge than a chief physician, and surpasses a real physician. The odds are high.
It is human initiative to ask questions and think about problems. How can we make computers learn to think about medical problems like a chief physician? This requires deconstructing medical knowledge into a language that computers can understand and use. For example, a drug can have 202 dimensions in "med-go", which are related to the symptoms of the disease, the patient's age, the patient's living environment, etc. No doctor can carefully think about using one drug from 202 dimensions. Plant medicine, but computers can, and computers can only think in 202 dimensions.
Zhang Haitao said: "Accurate and efficient explainable medical response content is the core competitiveness of medical models and the top priority in empowering medicine to improve clinical applications." "med-go" can effectively assist doctors in analyzing complex and rare problems. Cases, providing accurate and transparent clinical diagnosis and treatment recommendations.
“med-go” supports “medical education, research and management”
During the teaching rounds at Dongfang Hospital, teaching teacher Professor Xu Shumin was explaining cases to a group of intern doctors. In addition to giving students a detailed analysis of the diagnostic ideas, differential diagnosis and treatment options of the case, she also demonstrated how to use "med-go" for assisted learning - personalized explanation of knowledge points, case analysis and literature interpretation to help medical Students and residents can master professional knowledge more efficiently, and the system also supports the development of teaching plans and evaluation plans, making the teaching process more standardized and intelligent.
In the field of scientific research, “med-go” has also demonstrated strong strength. Dr. Liu Xiaobin of the extracardiac ICU is designing a study on the application of sglt2 inhibitors in patients with heart failure. "'med-go' can not only quickly analyze literature to find innovation points, but also provide professional evaluation and improvement suggestions, help optimize statistical plans, etc., which has greatly improved my scientific research efficiency." Dr. Liu Xiaobin said.
"med-go"'s strong medical professional capabilities and data processing capabilities also allow hospital management to further realize "data speaks". Taking medical record quality control management as an example, "med-go" was introduced into the doctor workstation of Oriental Hospital. The system automatically reviews and scores medical records based on a set of strict scoring standards, and gives specific scores and improvement suggestions for each item to help Doctors complete medical records and improve the quality of medical care. On this basis, Dongfang Hospital is actively planning to carry out comprehensive and efficient management of medical resources based on the "med-go" medical model and the construction of a "smart hospital" management system.
Xu Zhaohui, deputy director of Dongfang Hospital, said that the biggest feature of "med-go" is that it comes from doctors and serves doctors. As a large medical model entirely initiated by a team of doctors, with the main creator participating in the research and development, med-go’s clinical decision support capabilities have always been the core insistence on “excellence”. In the next step, the hospital will continue to rely on the joint laboratory to continue to improve system functions, including more in-depth clinical decision support, smarter medical quality management and more personalized medical education solutions, so that "med-go" can become a partner with doctors. A powerful tool for joint in-depth exploration in the medical field to truly realize the deep integration of artificial intelligence and medical care.