中国科研团队研发DeepRare系统破解罕见病确诊难题

Core Insights - The article highlights a groundbreaking achievement in AI rare disease diagnosis by a collaborative team from Shanghai Jiao Tong University and Xinhua Hospital, with the introduction of the DeepRare system, which addresses the global challenges of difficult diagnosis and high misdiagnosis rates in rare diseases [1] Group 1: DeepRare System Overview - DeepRare is the world's first intelligent evidence-based reasoning diagnostic system for rare diseases, developed to overcome the challenges of traditional medical AI [1] - The system employs an innovative "central-body" architecture that surpasses traditional medical AI across three dimensions [3] Group 2: Key Features of DeepRare - The system integrates vast medical literature and clinical case data, breaking down data silos to provide comprehensive knowledge support for each diagnosis [3] - DeepRare possesses a "slow thinking" capability akin to human doctors, allowing for iterative hypothesis testing and self-reflection to fill information gaps and correct logical flaws [3] - It features a transparent reasoning process, providing a complete evidence chain for each diagnostic conclusion, addressing the "trust crisis" in AI healthcare [3] Group 3: Performance Metrics - DeepRare achieved a first diagnosis accuracy rate of 57.18% using only patient clinical phenotypes, a 23.79 percentage point improvement over the best international models in the field [5] - In retrospective human-machine comparisons, DeepRare's diagnostic recall rate surpassed that of rare disease specialists with ten years of clinical experience, marking it as the first method to outperform human doctors in this metric [5] - When genetic data is included, the accuracy rate for complex cases exceeds 70.6%, significantly outperforming the widely used Exomiser tool, which has an accuracy of 53.2% [5] Group 4: Implementation and Future Plans - The DeepRare online diagnostic platform is set to launch in July 2025, having already attracted over 1,000 professional users and covering more than 600 medical and research institutions globally within six months [7] - The system is currently undergoing internal testing at Xinhua Hospital, where it will serve as a "digital quality control officer" for rare disease diagnosis [7] - The collaborative team is initiating a "10,000 clinical validation plan" to complete tens of thousands of real-world validations of difficult rare diseases within six months, aiming to extend this "Chinese solution" to global rare disease patients [7]