中国科研团队，研发出AI看病系统

Core Insights - The article discusses the development of DeepRare, an AI system for diagnosing rare diseases, which is the first of its kind to provide a traceable reasoning process for its diagnoses [1][3]. Group 1: System Features - DeepRare offers a complete "evidence chain" with each diagnosis, akin to how experienced doctors explain their reasoning during patient evaluations [3]. - The system utilizes a vast repository of global medical knowledge and real case studies, allowing it to hypothesize, validate evidence, and self-correct before reaching a conclusion [3]. Group 2: Performance Metrics - The initial diagnostic accuracy of DeepRare, based solely on clinical symptoms without genetic testing, is reported at 57.18%, which is an improvement of nearly 24 percentage points over previous best models [3]. - When genetic data is included, the diagnostic accuracy of the system exceeds 70% [3]. Group 3: Implementation and Adoption - Since its launch in July of the previous year, DeepRare has been adopted by over 600 medical institutions globally, ranging from major hospitals in China to leading laboratories in Europe and the U.S. [4]. - The system is already deployed at Shanghai Xinhua Hospital and is set to assist doctors in diagnosing rare diseases by filling in gaps and preventing missed diagnoses [4]. Group 4: Future Plans - The research team plans to establish a global alliance for rare disease diagnosis and treatment, aiming to validate the system further with 20,000 real cases in the next six months [4].