首个！中国团队研发出一款AI看病系统，成功破解世界难题

Core Insights - The article highlights the launch of DeepRare, the world's first intelligent rare disease diagnostic system developed by a joint team from Shanghai Jiao Tong University and Xinhua Hospital, addressing the global challenge of difficult diagnosis and high misdiagnosis rates in rare diseases [1][4]. Group 1: System Overview - DeepRare employs an innovative "central-body" architecture, surpassing traditional medical AI by breaking data silos and integrating vast amounts of medical literature and clinical cases for deep knowledge internalization [4]. - The system abandons traditional AI's pattern matching, utilizing a human-like "slow thinking" capability through "hypothesis-verification-self-reflection" to analyze diagnostic clues [4]. - DeepRare provides full-process evidence-based "white-box reasoning," ensuring that each diagnostic conclusion is accompanied by a complete traceable evidence chain, addressing the trust crisis in AI healthcare [4]. Group 2: Performance Metrics - Clinical data validation shows that DeepRare's pure phenotype diagnostic accuracy reached 57.18%, an improvement of 23.79 percentage points over the best international models, changing the paradigm that "without genetic testing, diagnosis is difficult" [4]. - The system's diagnostic recall rate surpasses that of rare disease specialists with ten years of clinical experience [4]. - After incorporating genetic sequencing data, the comprehensive first-diagnosis accuracy for complex cases exceeded 70.61%, significantly outperforming commonly used international tools [4]. - The inference reports generated by the system achieved a 95.40% satisfaction rate among doctors, ensuring that diagnoses are "well-founded" [4].