"Finally we have welcomed reinforcements." "It would be better if you came a few days earlier!" For nearly a month, Shi Lei, vice president of Itu Medical, has received feedback from frontline doctors in the epidemic.
This "reinforcement" is the industry's first intelligent assessment system for new coronavirus pneumonia jointly developed by Shanghai Public Health Clinical Center (Shanghai Public Health Center for short) and Itu Medical.
On the 30th of the new year, the project was formally established. On the fourth day of the first year, the first version was launched in the Shanghai Public Center. On February 5, it was quickly deployed to the Wuhan COVID-19 designated hospital within 24 hours.
From Wuhan University People's Hospital, Huazhong University of Science and Technology Tongji Medical College Affiliated Union Hospital, to Wuhan University Zhongnan Hospital, Jingzhou First People's Hospital ... At present, the system has been deployed in more than 20 provinces, municipalities and autonomous regions, including Hubei, Shanghai and Guangdong A number of medical and public institutions, this system uses artificial intelligence algorithms to play an important role in assisting doctors to quickly screen high-risk and suspected cases of CT images of patients' lungs, quantitative analysis, and evaluation of efficacy.
AI medical "raising soldiers" for more than four days
New products developed for the latest emerging diseases, from data collection, algorithm model building, and product verification, in the industry's view, it can take several months or a year. In this short time, the product was deployed on the front line of the hospital. How did you do it? "Rome was not built in a day."
Shi Lei said that the AI system developed for the epidemic is not a "prompt gun." "In terms of results, the four-day time is fast, but before that, our technological exploration in the field of AI medical treatment and the accumulation of knowledge in the medical industry were the key guarantees for rapid landing." After the epidemic, technology research and development It is based on the existing underlying architecture to adjust for the medical characteristics of COVID-19.
In August last year, the Ministry of Science and Technology announced that relying on Yitu Technology to create a "new generation of artificial intelligence open innovation platform for visual computing country", and Yitu has also accumulated relevant technology in the AI medical field for a long time. "It's like training a doctor," Shi Lei said. The doctor needs to use small samples of data to achieve rapid learning through algorithms when the amount and quality of case data is limited. On the other hand, in the face of some diseases, the AI "doctor" can also use common medical knowledge to find common problems from the underlying technology, and build a new algorithm underlying technology framework based on multi-modal medical data.
The characteristics of COVID-19 in CT images and many inflammations are intersected, and most of them are patchy. The first step in training AI is to make it recognize all inflammation-related features. And this can't be done in just four days.
As early as the second half of 2018, Yitu Medical began to build algorithmic models for lung multi-task intelligent diagnosis of chest CT. "Before the advent of COVID-19, general-purpose technology had been reserved." Shi Lei said that after the epidemic occurred, the R & D team according to Figure only needed to quickly find out what the differences and commonalities of COVID-19 were compared with traditional pneumonia. What kind of data or technical adjustments are needed to quickly sort out the diagnosis.
This spring, the heroes are fighting
"As soon as the phone rang, the team had already assembled; every day and night, dozens of working groups beating, recording every bit of this trip; every morning, we read through the last class of colleagues as fast as doctors' shifts in the last night. In the early morning, he turned around and started a new day. "In early February, the busy Shi Lei wrote in the company's internal WeChat account.
The R & D team is running at full capacity--remote collaboration, rapid awareness of new diseases, customer communication, product iteration, and algorithm accumulation are all tests. "We make full use of 24 hours a day, assemble more than a hundred people in various departments such as R & D, operation, scientific research and so on, and divide them into several groups, as far as possible, disassemble all capable things, group discussions, and advance in parallel." Shi Lei introduced.
After the product goes online, the work is far from over, the epidemic situation changes very fast, and the diagnosis and treatment plan is constantly iterating. They will listen to the needs and feedback of front-line doctors in time to iterate the product. The R & D team has many members with medical knowledge background, which allows them to fully understand medical problems during product development and iteration, quickly capture user needs, and continuously enrich and improve product functions.
Shi Lei, who has been a radiologist for many years, said that in the past, as a doctor, he would petition the frontline, and now he and his team support frontline medical staff through technological innovation.
Epidemic prevention and control requires scientific and technological innovation. The role of AI medical in epidemic prevention and control goes far beyond medical image-assisted diagnosis. Epidemic prevention knowledge education and publicity can cut off high-risk groups at an early stage. The image-assisted diagnosis system screens suspected cases to achieve early diagnosis, early treatment, early isolation, quantitative evaluation of patient treatment effects during treatment, and subsequent follow-up of new epidemic stages ... … Undertake the medical data market behind the epidemic, build an integrated intelligent prevention and control system, and Yitu Medical has already completed its technical reserves and constructed solutions.
Yitu Medical also developed the "COVID-19" intelligent identification service for the epidemic situation—Yitu new-type coronavirus epidemic prevention doctor Xiaoyi, using WeChat public account, mini-program, etc., to ease the pressure of medical institutions; In addition, efforts to strengthen prevention and control and resume In terms of resumption, Yitu also launched a comprehensive solution called "cloud edge", which integrates thermal imaging temperature measurement technology, whether to wear a mask detection, and face recognition capabilities.
"This epidemic is a test. Let us see that people with various knowledge backgrounds are working hard to make corresponding contributions to the epidemic. The current epidemic has entered a new stage. More behind-the-scenes heroes are still needed to finally seize the epidemic. Victory. "Shi Lei said.
0 Comments