The intelligent diagnostic system already "reads" the medical records

artificial intelligence diagnostic system can not only "see the picture", but also "literate". A team of researchers from Professor Xia Huimin of Guangzhou Women's and Children's Medical Center, Itu Medical, Kangrui Intelligent Technology and Guangdong Province's Key Laboratory of Regenerative Medicine has developed an artificial intelligence diagnostic system for pediatric diseases. After learning the diagnostic logic of 1.36 million high-quality electronic text medical records of 567,000 children, the system can read the medical records, recommend the diagnosis, and diagnose the common pediatric disease with 88.5% accuracy. The paper was published online February 12 in the international medical research journal Nature Medicine.recent years, the use of artificial intelligence diagnostic tools has been limited to relatively standardized still image data reading, said Xia Huimin, a researcher at the University of China. The system developed by the team can further understand and analyze complex cases by automatically learning the diagnostic logic in the medical record text data. A team of more than 30 senior pediatricians and more than 10 informational researchers manually annotated more than 6,000 charts on electronic medical records and continuously examined and iterated the model, while using a medical history intelligent analysis system established using natural language processing techniques. The unstructured text form of medical records data into structured data, so that the artificial intelligence system can accurately and completely read the medical records, including patient complaints, symptoms, personal history, physical examination, laboratory test results, imaging results, drug information and other aspects of the data.researchers randomly selected 12,000 children's medical records and divided 20 "participating" pediatricians into five groups based on seniority and clinical experience. The results showed that the accuracy of diagnosis in AI systems was 88.5%, higher than that of two groups of junior doctors (84.1% and 83.9%).team has also developed an intelligent recommendation system for diagnostic results. The system is first divided into respiratory diseases, gastrointestinal diseases, systemic diseases, and so on, and then broken down under each category. A primary diagnosis made by artificial intelligence with near-precision that is close to the primary diagnosis made by a physician. (Health Report)