Reducing the repetitive work in cancer screening is the mission of AI

Original title: board a set of CCTV's "super wit" and think deeply about AI's "Qiaosi" Zhijing pathologist Author: Yang xueevery two minutes in the world, one woman died of cervical cancer The best way to snipe at cervical cancer, the second most lethal cancer in women, is to treat it before its precancerous lesions The cure rate of cervical cancer can be as high as 90% by early screening But patients with advanced cervical cancer have a 3% chance of survival It can be said that regular early screening is the boundary between life and death of cervical cancer patients However, in China, the screening rate of women's cervical cancer is still far lower than the target requirement of 80% coverage rate required by the women's development program (2011-2020) For example, in the grass-roots areas such as Shangrao, Jiangxi Province, about 2.2 million of the 3.31 million permanent female population need cervical cancer screening, but there are only five doctors with screening ability in the city He Weihua, a pathologist from Shangrao basic hospital in Jiangxi Province, lamented: "if all the early screening tasks are completed by doctors, it will take about 70 years to complete the screening tasks Obviously, the main reason lies in the shortage of pathologists, and it usually takes more than 8 years to cultivate a new pathologist Artificial intelligence may be able to help people twist the balance of life and death again, and improve the early screening and early diagnosis rate of cervical cancer In a set of the latest CCTV program "smart beyond people", 10 pathologists from China's top three hospitals, such as Peking Union Hospital, Beijing Friendship Hospital, affiliated obstetrics and Gynecology Hospital of Zhejiang University Medical College, Peking University People's Hospital, were gathered on site to test the artificial intelligence cervical cancer assistant screening robot - Qiaosi In the program, with the help of ideewise.ai-qiaosi robot, Dr he Weihua competed with 10 senior pathologists to judge 30 cervical cell samples and test the screening ability of artificial intelligence The field results show that through the ideewise.ai-qiaosi robot, the grass-roots doctors can achieve "one for ten" and give the diagnosis results completely consistent with the senior pathologists in the same time Among them, ideewise.ai-qiaosi robot comes from a company called "deep thinking artificial intelligence" Let doctors return to the diagnosis work, and precipitate the wisdom of doctors The theme of this program is "wisdom" to the pathologist In addition, please invite Bu Hong, the executive director of China Anti Cancer Association and the chairman of China Society of Pathology, Professor of pathology of West China Hospital of Sichuan University, as the "wisdom" to see the witness As the first batch of pathologists after the reform and opening up, Professor Bu Hong witnessed the important role of improving the quality of cytopathological screening for early cancer screening Professor Bu Hong said: "the pathology department serves all clinical departments of the hospital and is known as the doctor's doctor Pathology is closely related to tumor prevention For a long time in the past, we can't talk about the regular early screening of tumors Only when we suspect that there is a tumor can we screen it Today's medical technology allows us to detect cancer early, but it takes a lot of time for pathologists to observe and identify it through a microscope " Cervical cancer is divided into cervical squamous cell carcinoma and adenocarcinoma In the traditional cervical cancer cell image screening, doctors use a microscope to observe the cell smear, and need to constantly switch three times of the lens First, use a 100 times of the whole piece to observe carefully, and then find the abnormal cells to change 200 times or even 400 times of the whole piece to observe carefully There are tens of thousands of cervical cells on the slide less than one inch Doctors first need to screen tens of thousands of normal cells, find the diseased cells, and then diagnose according to cell morphology and combination To find out the diseased cells in tens of thousands of cells, the work of pathologists is like looking for a needle in a haystack, while the AI system can be said to send the "needle" to the doctor Ideewise.ai system can eliminate the negative cells one by one, and show the questionable cells in front of doctors, so that doctors can easily judge whether the cells are pathological changes and shorten a lot of screening time Let doctors return to the essence of diagnosis, instead of spending a lot of time to find and screen negative cells, to alleviate the shortage of existing pathologists In China, there are 100 million women who need cervical cancer screening, but there are only about 10000 licensed pathologists in China For a single pathologist, artificial intelligence reduces the workload For the whole group of pathologists, artificial intelligence precipitates the wisdom of doctors and reduces missed diagnosis and misdiagnosis Professor Bu Hong said: "pathological diagnosis relies on the experience of doctors, but pathologists have a high workload There is a saying in our industry that when a pathologist has enough experience, it's time for him to have cataracts But artificial intelligence can learn and accumulate constantly It is a doctor who will not die " The original wisdom of AI also comes from doctors The foundation of ideepwise.ai Qiaosi system learning comes from the deep thinking 1.2 million cervical cell sample database This data center grows nearly 1000 samples every day Deep thinking AI has not only learned millions of high-quality sample data, but also carried out a lot of AI core level algorithm innovation It adapts a variety of production methods and scanning equipment, realizes full-automatic cervical cytology assisted screening through the integration of deep learning, image instance segmentation, migration learning, visual semantic understanding and other artificial intelligence technologies, which can be used for TCT (membrane), LCT (sedimentation) TLT (centrifugation) and other liquid-based cytology slides were read to provide reliable test results Taking the world's most authoritative cervical cell image Herlev data set as an example, the best result in this data set is the latest paper issued by NIH, the National Research Institute of the United States Deep thinking about the AI team's innovative AI algorithm, in the same data set, under the same evaluation criteria, the above results are comprehensively exceeded, among which the accuracy of cell classification in key indicators reaches 99.3%, achieving this goal The best result in the world in the dataset Through a large number of clinical samples validation, ideepwise.aicias system significantly improved the detection sensitivity of ascus and above lesions, significantly reduced the false negative rate, and greatly reduced the missed diagnosis Among them, the sensitivity of squamous epithelial cell lesions was 99.4%, the specificity was 98.9%, the sensitivity of glandular cell abnormalities was 97.9%, the specificity was 93.4%, the negative discharge rate was 81.19%, and the false negative rate was as low as 0.006% It is understood that the deep thinking artificial intelligence assisted screening system for cervical cancer is the only AI product that can effectively identify the abnormalities of glandular cells in China Breaking through the technical bottleneck, accelerating the landing of the product on the scene of "super smart", said Dr Meng Zhilan from the pathology department of Peking Union Medical College Hospital: "I think the most difficult thing is the cells stacked together, i.e clustered cells We can only adjust the microscope to a high power, and constantly adjust the fine spiral to see every cell as well as possible If it still cannot be determined, it is necessary to classify it as suspicious lesions and make further biopsy diagnosis " In view of the "difficulty" in reading the film, ideepwise.ai initiated the deep learning segmentation algorithm of cluster cells Dr Yang Zhiming, founder of deep thinking, said: "our AI clustering cell segmentation algorithm can separate cell images from clusters and then interpret them." In different hospitals and institutions, there are great differences in the stained cell images, which directly reduces the efficiency of pathological doctors' reading The depth learning algorithm is applied to the dye standardization of digital pathological images, which realizes the automatic adjustment of the dye depth of unconventional dyed cell images In addition to this series of pioneering algorithms, I think there are many breakthroughs in AI medical kernel technology It is worth mentioning that one of the 30 so-called cervical cancer cell images on the scene of wit is urine sediment When recognizing this image, Dr Yu Xiaomeng, a 30-year-old pathologist from Beijing Friendship Hospital, recognized the problem, and ideewise.ai showed no inferior performance, quickly identifying abnormal cells Enabling cervical cancer screening at grass-roots level with weak medical foundation through AI is also the direction vigorously promoted by the health care Commission In the maternal and child health indicators of Health China action 2019-2030, the health commission plans to promote the cervical cancer and breast cancer screening coverage of rural women of appropriate age to 80% and above and 90% and above respectively by 2020 and 2030 The health and Health Commission also made it clear that it would use AI and other means to increase the screening coverage rate of "two cancers" (breast cancer and cervical cancer) by 27% within three years Deep thinking: while increasing investment to break through the bottleneck of AI cervical cancer screening technology and improve the technical height, it is also focusing on the rapid application of products With the support of the world's first special AI chip for medical treatment and the AI algorithm model of pathological cytology independently researched and developed, the microscope and pathological scanner are "equipped" with the terminal and the "artificial intelligence brain" in the cloud, which can simultaneously process 480 smears and complete 70000 cell interpretation of a single smear in 60 seconds Its application is very suitable for the situation of the country, and it is beneficial to the grass-roots hospitals to solve the problem of unbalanced medical resources It is understood that deep thinking AI cervical cancer screening system has been applied in hundreds of hospitals, including more than 30 top three first-class hospitals, including Shanghai Fudan University Cancer Hospital, Zhejiang University Affiliated First Hospital and other hospitals, providing AI cervical cancer early screening services for thousands of users Returning to the scene of "smart beyond people", Dr Yang Zhiming firmly said: "our vision is to use artificial intelligence technology to turn the balance of life and death, and" smart "respects life." Professor Bu Hong, the witness, said: "we usually have two tests to judge whether a technology can work The first is to test its ability The second is to get the examination and approval certificate in the national testing center before it can go to work I think (think deeply of ideewise AI) the smart thinking robot can have the ability to work " CCTV's "super smart" -- cervical cancer screening system enables cancer prevention in the future program video This article is reprinted by yaozhi.com, and the copyright belongs to the original author The purpose of reprinting is to transmit more information, which does not represent the view of this platform If the content of the work, copyright and other issues are involved, please contact our website message, we will delete the content in the first time.