Core reading Looking at the film, who is stronger in the human brain? Recently, a "human-machine war" in the field of neuroimaging ended in Beijing. Medical artificial intelligence beats the “human team†of many neurological experts with faster speed and higher accuracy. Experts say that artificial intelligence is expected to free doctors from the diagnosis of heavy readings, have more time to do research, warm patients, and save patients' cost of seeing a doctor. At the same time, experts also reminded that image interpretation is not just a film, but also a history of the patient. Different diseases sometimes have the same performance, requiring doctors to comprehensively judge through years of accumulation. The "human-machine war" in the field of neuro-images, which has received much attention, has recently come to an end in Beijing. The competition was jointly sponsored by the National Center for Neurological Disease Clinical Medical Research, the Capital Medical University Human Brain Protection Center and the China Stroke Association. The artificial intelligence (AI) player "BioMindTM" has defeated the "most powerful brain" of 25 medical circles from the global diagnosis of neurological diseases with a 20% accuracy rate. What changes will medical AI bring to brain medicine? Competing speed computer faster When the countdown on the big screen in the field is 14 minutes and 30 seconds, the prompt for answering the question is sounded. Hundreds of viewers on the scene could not help but scream. It turned out that the "BioMind Tianzhizhi", which had to complete 225 questions independently, had been handed over for nearly 15 minutes, and the "human team" was still nervous. The official competition is divided into two groups of AB. The party in Group A is a member of 15 “Human Team†doctors. It consists of 6 winners from the national online preliminaries, 7 experts in the forefront of domestic neurological diseases, and 2 well-known foreign hospital experts. Each doctor needs to interpret CT and MRI images of 15 intracranial tumors and hematoma prediction within 30 minutes. The other side of the artificial intelligence BioMind also needs to complete the work of 15 doctors in 30 minutes, that is, complete a total of 225 questions. There are 10 doctors in group B, all of whom are above the deputy director level. They are enrolled in CT and MRI images of cerebrovascular diseases. Each doctor in group B needs to read 3 films in 30 minutes, allowing doctors to discuss the results of the interpretation. BioMind needs to read 30 films in 30 minutes. In front of the "Human Team" players are a computer showing the title, a paper version of the answer sheet and a pen. When answering the questions, the players are staring at the screen, dialing the mouse to view the image, and thinking about the chin from time to time. . On the other side, on the screen of “BioMind Tianzhizhiâ€, swiftly swept a film and quickly showed the brain tumor disease reflected by each group of films. The artificial intelligence machine participating in the competition was developed by Beijing Tiantan Hospital and Beijing Ande Medical Technology Co., Ltd. After teaching Professor Gao Peiyi, director of the Neuroimaging Center of Beijing Tiantan Hospital, “BioMind Tianzhizhi†has become the most “young†disciple of the neuroimaging team of Tiantan Hospital. Jing Lina, deputy director of the Department of Radiology at Beijing Tiantan Hospital, is a 14th player. Her answer is the same as usual. One or two of the 15 questions are not accurate. If this happens, she will write two answers, and then ask the senior experts for discussion, or discuss them together. I can only write one answer during the game, and I thought about it for a long time. Massive sample mastery In the end, is the "best brain" to win, or is this medical AI that just came out to win? In this competition, the group A test questions were randomly selected from the Tiantan Hospital brain tumor case base. Group B was the Beijing Tiantan Hospital executive vice president Wang Yongjun randomly selected from the National Neurological Disease Clinical Medical Research Center cerebral hemorrhage case database, two The group questions are not AI training questions. In order to ensure the confidentiality of the questions, the notary from Beijing Chang'an Notary Office has notarized the selection and sealing of the questions. The results of the final live competition were: 225 cases in Group A, 87% accuracy in AI, 87% in 15 minutes, and 66% accuracy in 15 doctors in Group A; Group B: AI used 15 minutes with an accuracy rate of 83%, and 10 doctors used 30%. The minute accuracy rate is 63%, and the AI ​​wins regardless of time or accuracy. Jing Lina is psychologically prepared for its speed. The case analysis of Tiantan Hospital has been recorded in its system for many years. The sample size is super large, and it is instantaneous, and the speed is definitely better than it. But I really didn't expect to lose in terms of accuracy. In the face of the two rounds of "BioMind Tianzhizhi", its teacher Gao Peiyi said that through the systematic study of tens of thousands of cases of nervous system-related diseases that Beijing Tiantan Hospital has received in the past ten years, "BioMind Tianzhizhi" The diagnostic ability of magnetic resonance imaging in common diseases such as meningioma and glioma is equivalent to the level of a senior professional physician, and the strength cannot be underestimated. Behind each type of tumor, it has studied 1000 cases, and currently has basically mastered 50 kinds of brain tumors, which is difficult for any doctor to achieve. "I am not satisfied with its performance, I think its accuracy should be above 90%." Gao Peiyi said, they will then analyze the reasons for the "distribution" of AI. “The machine training time is not enough. If you can give AI more learning time, it will perform better.†Wang Yongjun said that if the accuracy of AI prediction for hematoma can exceed 85%, the National Center for Neurological Disease Clinical Medical Research plans. Start a clinical validation study. How is the prospect of artificial intelligence? The speed is high, the accuracy is high, and the medical AI performance is amazing. At the same time, more people are concerned about what changes it will bring to neuromedicine in the future. Can human brain diseases achieve computer diagnosis? Gao Peiyi introduced that an imaging doctor at Tiantan Hospital had a diagnosis time of 18 hours per day. According to the current speed of "BioMind Tianzhizhi", a doctor's daily workload, it only takes 400-500 seconds, which is less than 10 minutes. Now the patient goes to the hospital to do nuclear magnetics, and the results will have to wait until the next day. If you let the AI ​​do it, then the nuclear magnetic results are basically equal, just a few minutes. "Doctors work very hard. I hope that in the future, AI will free the doctors, let doctors have more time to do research, warm the patients, and save patients the cost of seeing a doctor." Gao Peiyi said. For the rapid development and amazing achievements of “BioMind Tianzhizhi†in the field of neuroimaging-assisted diagnosis, Wang Yongjun said that it has continuously improved the efficiency and accuracy of disease diagnosis in just a few months; many hospitals can be learned for many years. Rare and difficult cases that are not seen, have innate advantages and big data foundation in the research and development and learning depth of the neurological field. At present, the application has been submitted to the State Food and Drug Administration, and it is hoped that the AI ​​product will be applied in the clinic to improve the accuracy of image diagnosis in primary hospitals, and also improve the speed of image interpretation, saving patients time. Professor Xie Peng, one of the judges of this final and director of the Neuroscience Center of Chongqing Medical University, said: "For the diagnosis of existing knowledge, from the perspective of deep learning of big data, the probability of AI winning is much greater. It should be clear to the chest. But in some new and specific diseases, there is still not much consensus in the medical field, AI may still 'make' but human." "I don't think it will win me in the future." Jing Lina said that it does not understand the complexity of medicine. Image interpretation is not just a film. It also has an understanding of the patient's medical history, including laboratory tests. Different diseases will have the same performance. It is necessary for doctors to judge which one they prefer through years of accumulation, rather than simply relying on nuclear magnetic films and through certain signs. "I personally don't care very much about who wins and loses this game." Wang Yongjun said that this neuroimaging artificial intelligence man-machine contest is not intended to provoke a war between human doctors and AI, hoping to pass this game and allow Doctors have experienced the charm of artificial intelligence, especially for some skeptical doctors to further understand artificial intelligence. “Standard labeling of a large number of cases is both an opportunity for AI development and a bottleneck restricting development. The real AI technology is a process of continuous learning and will never be perfect.†Wang Yongjun said that the AI ​​technology with “Tiantan Standard†is gradually maturing. Later, it will become an auxiliary grassroots doctor, especially how the grassroots doctors in remote areas can read, diagnose and predict the learning and training tools of the film, so that patients with encephalopathy can obtain high-quality and personalized diagnosis and treatment programs at the “home entranceâ€. In addition, it can help doctors complete the screening and assessment, and finally the doctor will confirm the judgment, improve the work efficiency, save the repetitive mechanical work, especially when the judgment results are inconsistent, the doctor can be reminded to avoid misdiagnosis. People's Daily (13th edition, July 2, 2018) Dehydrated Coriander,Dehydrated Coriander For Seasoning,Dehydrated Coriander For Cooking,Dehydrated Coriander Powder Xinghua Lvwei Foods Co.,Ltd , https://www.lvweifoods.com