A few days ago, at the Machine Learning for Preventing and Combating Pandemics (MLPCP) Global Challenge initiated by ICLR 2021, Tsinghua-Xunfei Joint Research Center took part in the Disease Automated Diagnostic Dialog System The Tsinghua-Xunfei Joint Research Center participated in the Dialog System for Automated Disease Diagnosis (DADD) track and won the first place among the 73 teams from all over the world.
The International Conference on Learning Representations (ICLR) is one of the three most prestigious top-level international conferences in the field of machine learning. In order to promote the development of machine learning technology in the field of healthcare, ICLR 2021 launched this competition. With the development of automatic medical diagnosis dialog system as the theme, this competition is the first international challenge event for medical dialog generation and diagnosis, with two major tracks: medical dialog generation track and automatic disease diagnosis dialog system track. In the end, Xunfei Medical successfully won the champion of the automatic medical diagnosis dialog system track.
Xunfei Medical's success in winning the championship is based on the word "accurate". Its automated medical diagnostic dialog system is able to simulate a dialog between a doctor and a patient in a textual way. First, the system predicts the disease based on the patient's self-reported symptoms; then, the system predicts the next symptom to be asked based on the prediction result, and so on until the stop condition is met; finally, all the symptoms obtained will be sent to the automatic diagnosis system for disease prediction.
In medical diagnostics, the accuracy of "asking" and "diagnosing" directly determines the effect of diagnostics. During the competition, Xunfei technology passed the test of 12 diseases, 118 symptoms, and a total of 2374 conversation data, and finally won the championship in the global 73 teams, and the diagnostic accuracy rate realized the "world's first".
As the world's leading artificial intelligence company, KU Xunfei has made remarkable achievements in the field of smart healthcare in recent years. Through the construction of artificial intelligence assisted diagnosis and treatment system, KU Xunfei's AI technology y cut into the doctor's clinical diagnosis process, in the doctor's diagnostic process to give auxiliary diagnosis advice, to enhance the doctor's diagnostic and treatment ability and service level, especially the grassroots doctors.
At present, KDDI A.I. technology has achieved remarkable results in regulating primary diagnostic and treatment behaviors, enhancing the level of primary medical services and other livelihood areas. According to the Xunfei 2020 financial report disclosure, Xunfei intelligent medical assistant based on the "depth model + knowledge reasoning" auxiliary diagnosis and treatment technology innovation, for the grass-roots medical personnel to provide medical records quality control, assisted diagnosis, rational use of medication, medical knowledge retrieval and other functions have been better applied. 2020, Xunfei intelligent medical assistant to achieve the Anhui Province, grass-roots medical institutions comprehensive coverage, and in the western part of the country, the Xunfei A.I. technology has been used to enhance the level of health care services. In 2020, KDDI will realize the full coverage of primary medical institutions in Anhui Province, and gradually start the application in Tibet, Inner Mongolia, Qinghai, Beijing, Xinjiang, Heilongjiang, Zhejiang, etc. It has already been on-line in more than 30,000 primary medical institutions in more than 170 districts and counties across the country, and serves tens of thousands of primary medical practitioners.
At this stage, there is still a relative lack of high-quality medical resources and a structural problem of insufficient primary medical service capacity in China. We believe that as the scale of smart healthcare applications continues to expand, our quality healthcare resources will be more equitably and efficiently distributed.