罕见病有了智能可信的AI“听诊器”

Core Insights - Shanghai Jiao Tong University has developed the world's first intelligent rare disease diagnostic system, DeepRare, addressing the global challenge of high misdiagnosis and underdiagnosis rates in rare diseases [1][2] - The system surpasses traditional medical AI by integrating vast medical literature and real clinical case data, enabling deep understanding and internalization of medical knowledge [1] - DeepRare demonstrates significant improvements in diagnostic accuracy, achieving a first-position accuracy of 57.18% without genetic data, and over 70.6% with the inclusion of genetic sequencing data [2] Group 1 - DeepRare's ability to perform "phenotype decoding" has improved the diagnostic accuracy by 23.79 percentage points compared to previous best models in the field [2] - The system's reasoning reports, which include a complete evidence chain, have received a 95.4% approval rate from human experts, ensuring that diagnoses are well-supported [2] - The platform has been operational since July 26, 2025, and has attracted over 1,000 professional users from more than 600 medical and research institutions globally [2] Group 2 - The research is a result of a long-term "medical-engineering intersection" strategy at Shanghai Jiao Tong University, led by a collaborative team of professors from the AI and medical schools [3] - The team plans to conduct real-world clinical validation of thousands of rare disease cases over the next six months through an international multi-center cooperation network [3] - The project aims to create a global intelligent diagnostic network to shorten the diagnostic journey for rare disease patients [3]