Two units obtained the first AI Class III medical device registration certificate in China's in vitro diagnostic industry!

According to the WeChat public account of Peking Union Medical College Hospital on November 21, the peripheral blood cell image leukocyte assisted recognition software jointly developed by the Laboratory Department of Peking Union Medical College Hospital and Beijing Xiaofly Technology passed the approval of the National Medical Products Administration and obtained the first AI Class III medical device registration certificate in the in vitro diagnostic industry.

At present, most medical institutions recognize abnormal cells and normal cells through the human eye. Replacing heavy manual labor with artificial intelligence will help improve the quality and safety of medical care and bring a leap forward in improving the ability to diagnose diseases. Xu Yingchun, director of the laboratory department of Peking Union Medical College Hospital, said in an interview with People's Daily Health Client.

It is reported that complete blood cell analysis is one of the three basic tests of medical testing, and most of the disease diagnosis and treatment process requires complete blood cell analysis, of which peripheral blood cell morphology test is an indispensable and important means in complete blood cell analysis.

"In the past, the laboratory technicians use a microscope to see the cell morphology, which put forward high requirements for the experience and ability, and the test results will vary between different doctors. Through the core algorithm of cell recognition, the morphological feature extraction of winter cells can be effectively carried out to help clinical examination and avoid missed and mis detected as much as possible. Lu Dongqi, co-founder of Beijing Xiaofly Technology and head of the registration quality control department, told the People's Daily Health Client

"There is a problem with the patient's blood film, some may not be seen before, but now based on the characteristics of AI technology, the feature extraction of various cell forms is effectively carried out, and the recognition accuracy of normal cells can reach more than 99%, and the identification of abnormal cells can reach 95%." Wu Wei, deputy chief technician of the laboratory department of Peking Union Medical College Hospital, told the People's Daily Health Client. The accuracy of peripheral blood cell image white blood cell assisted recognition software is so high, it is inseparable from many preparatory works in the early stage of the team, facing the difficulty of massive data screening and labeling, each cell in the group is cross-labeled, repeated verification to ensure the high-quality output of data.