The robot “intelligent physician assistant†has scored high scores in the medical examinations. After the application in the future, it will help doctors improve the diagnostic accuracy and shorten the diagnosis and treatment cycle.
According to the Beijing News, “The four-hour 150-minute exam takes 15 minutes and 8 seconds, scoring 456 points and passing the 96-point pass.†On November 14, “Xiao Medical†received clinical practitioners in 2017. Comprehensive written test transcripts, data show that in the country in the same period of 530,000 candidates, this score can be ranked in the top.
“Xiao Medical†is a robot. The official name is “Intelligent Medical Assistantâ€. Developed by Tsinghua-Xinfei Joint Lab in February this year, it can realize autonomous learning, autonomous reasoning, and autonomous judgment. Wu He, director of the joint lab, said that the "intellectual assistant" did not benefit from the "sea tactics" through examinations, but instead developed for the needs of medical examinations to enable them to have multi-level reasoning abilities between vocabulary, sentences, and paragraphs. , will help improve the doctor's diagnostic accuracy, shorten the diagnosis and treatment cycle. It said that the robot plays an auxiliary role in diagnosis and treatment and will not try to replace doctors for independent medical treatment, and it will not replace people.
150 minutes robot time does not exceed 20 minutes
On August 26, 2017, the first day of comprehensive written test for clinical practitioners nationwide. On this day, 530,000 medical students and medical practitioners took the exam.
A special "candidate" also ushered in a challenge on this day. As an important project of the Tsinghua-Xinfei Joint Laboratory (hereinafter referred to as the Joint Laboratory), the artificial intelligence system “Intelligent Medical Assistant†participated in the period under the supervision of the National Health and Family Planning Commission, Anhui Provincial Health Planning Commission, and the National Medical Examination Center. Two days, four written tests before and after.
The Beijing News reporter learned from the joint laboratory that the “intellectual assistant†is in the examination room designated by the National Medical Examination Center. The answer process is notarized by the Beijing Guoxin Notary Office and is “strictly notarized and independent.†Examination room, off screen, on-site supervision." This is the first time that artificial intelligence robots and human candidates have participated in the "national" professional qualification examination on the global scale, and the artificial intelligence technology has been tested in the medical field for the first time.
In two days, the time required for each exam is 150 minutes. The actual answer time of the “intellectual assistant†robot is 17 minutes 36 seconds, 13 minutes 46 seconds, 5 minutes 09 seconds, and 16 minutes 44 seconds, averaging 15 minutes per game. Zero 8 seconds.
On November 14th, the Joint Lab announced to the public that after being reviewed by experts, “Intelligent Medical Assistants†had a total score of 456 points, ranking among national candidates at intermediate and advanced levels, and scores exceeded 90% or more of candidates in Anhui Province this year. This is also the first time that robots have passed a comprehensive written examination in the medical field.
Previously there was a constant precipitation of technology
The “intellectual assistant†has a total of 9 months before and after the project has passed the medical examination. On February 14th this year, Tsinghua University's multimedia signal and intelligence information processing laboratory, together with the AI ​​Research Institute of Science and Technology, and Xunfei Medical Company jointly established a research and development team for the medical examination system. The joint laboratory is responsible for core algorithm research and core engine development.
This is not the first time that artificial intelligence technology has entered the medical field. In an interview with local media, Xu Geliang, the director of the Anhui Provincial Hospital, introduced that in June 2016, Anhui Provincial Hospital cooperated with HKUST News and developed the artificial intelligence medical image-assisted diagnosis system. In the past year, the system has studied 680,000 CT images of the lungs and assisted doctors in the Provincial Hospital CT room to diagnose about 11,000 CT images with a diagnostic accuracy of 94%.
Since then, the Anhui Provincial Smart Hospital, which relies on artificial intelligence technology, was established. This is also the first in China. In the hospital, intelligent guidance medical robots, artificial intelligence medical imaging auxiliary diagnostic systems and other artificial intelligence devices have entered the application stage.
Even so, for artificial intelligence, it is technically difficult to pass comprehensive medical tests. The reporter learned that when the "Gentlemen's Assistant" first-generation products did mock test questions, they scored only 100 points, and they learned about 1 million medical image data, 53 professional medical textbooks, 2 million to identify the real electronic medical records, 400,000. The medical literature and medical record report achieved high scores after continuous iteration and research and development.
Decryption
Has a massive knowledge base for comprehensive analysis
Robots passed a written test by practicing doctors and their performance was higher than that of most candidates, raising concerns for a time. One view is that the secret of "intellectual physician assistants" is the "sea tactics" and depends on the computer's powerful memory. In this regard, the joint laboratory director Wu and denied.
Wu and introduced that the written test questions vary from year to year, so for the machine, the annual written test is "new". In addition, pen questions can be divided into two types: “knowledge type†and “case typeâ€. Since 2013, the new syllabus has required an increase in the assessment of clinical thinking ability, leading to a continuous increase in the proportion of “case-type†test questions. It has reached more than 50%. Therefore, simply relying on memory skills and rapid retrieval skills will obviously not work. It said that the "intellectual physician assistant" passed the medical examination because it possessed autonomous learning, autonomous reasoning, and autonomous judgment.
The Beijing News reporter learned that since the artificial intelligence technology was first applied to the medical examination field, the R&D team had no experience to follow. In the early stages of research and development, a problem arises: medical knowledge is large and complex, and traditional knowledge mapping technology cannot meet the needs of describing medical knowledge. It is even harder to make machine learning and know how to use it.
In response to this situation, Wu He said that the joint laboratory proposed a set of methods called “semantic tensor,†which includes inputting large amounts of data including medical undergraduate textbooks, clinical guidelines, and classic cases for medical examination needs. Make it have a huge medical knowledge base and achieve "some learning." In addition, through the "key point semantic reasoning", "context semantic reasoning", "evidence chain semantic reasoning" and other algorithms to enhance the machine's text reasoning ability, so that it will "learn".
In other words, after the “Intelligent Medical Assistant†obtained a professional knowledge base, it had the ability to think deeply and comprehensively through technological innovation, and finally completed medical knowledge testing. In this process, the machine will complete the process from reasoning to analysis and judgment. Wu and introduced that when answering questions, the robot first analyzed the questions and individual options, and searched for each option in the knowledge base as possible conditions for correct answers. Then analyzed the relationship between the questions, options, and evidence. Final judgment.
Extension
Developers: Artificial intelligence will not replace humans
The Beijing News reporter was informed that the “intellectual assistant†will be promoted throughout the country in March next year. Since it only passed the written test and did not participate in the practical skills test, it is still unable to obtain the qualification certificate of the practicing doctor.
Is artificial intelligence technology likely to replace humans? Joint laboratory gives a negative answer. One of the project leaders, Tao Xiaodong, the general manager of HKUST Medical News, publicly stated that practicing medicine alone would not be the direction of “intellectual physician assistant†development. It said that there are often unexpected situations in the actual diagnosis and treatment process, artificial intelligence technology currently only masters professional knowledge, and can not fully realize the application of knowledge. Wu He also stated that robots can play an auxiliary role in diagnosis and treatment, but they will not try to replace doctors for independent medical treatment, and they will not replace people.
According to the plan of the joint laboratory, after the “intellectual assistant†is applied, it will be used for medical education and standardized training. In addition, it will also assist the doctor in clinical diagnosis and treatment, which will help improve the diagnostic accuracy of the doctor and shorten the cycle of diagnosis and treatment. To the "medical secretary" role. For example, the traditional reading medical records and admission records work are not technically demanding, but they require a lot of energy for the doctors. The robots can do it in the future, and they can give preliminary diagnosis suggestions by understanding and extracting information. In addition, in the follow-up observations, "intellectual assistant" can also provide more medical advice for doctors to choose.
"Help doctors with 60 points to do 80 points, and doctors with 90 points to do 60 points less." At present, the "intellectual assistant" project has cooperated with the National Medical Examination Center to jointly explore the field of artificial intelligence in medical examinations. Further application.
Oil-immersed Distribution Transformer, its HV level is 15kV, LV is 400V, its capacity is 2500kVA and below. Generally installed on the pole or in the distribution room, for lighting or power supply, three-phase power supply, fully sealed tank structure, to ensure the transformer safe and reliable operation.
Generator Transformer,15Kv Distribution Transformer,15Kv Oil Immersed Transformer,High Quality 1000Kva Transformer
Hangzhou Qiantang River Electric Group Co., Ltd.(QRE) , https://www.qretransformer.com