SHANGHAI, Feb. 20 (Xinhua) — A Chinese research team has unveiled an AI-powered diagnostic system for rare diseases, known as DeepRare, which has set a new benchmark in diagnostic accuracy, as detailed in a study published Thursday in the journal Nature. This innovative tool addresses longstanding challenges in diagnosing rare diseases, especially in areas with limited access to genetic testing.
Traditional medical AI diagnostic systems have often struggled with trust issues, largely due to their opaque reasoning processes. In contrast, the evidence-based DeepRare was developed by a collaborative team from Xinhua Hospital, associated with both the Shanghai Jiao Tong University (SJTU) School of Medicine and the SJTU School of Artificial Intelligence. Since the launch of its online diagnostic platform last July, DeepRare has attracted over 1,000 professional users from more than 600 medical and research institutions globally.
Clinical test data reveal impressive results: when provided solely with patients’ clinical phenotypic information—excluding genetic data—DeepRare achieved a first-attempt accuracy rate of 57.18 percent in phenotypic diagnosis, representing an enhancement of nearly 24 percentage points compared to the previous global model. When genetic data was included, the diagnostic accuracy surpassed 70 percent.
The study highlights that DeepRare utilizes real-time access to a comprehensive repository of medical literature and real-world clinical case data. This system employs an iterative cycle of hypothesis generation, verification, and self-reflection, allowing it to evaluate diagnostic clues and address logical gaps in reasoning effectively.
Significantly, each diagnostic conclusion generated by DeepRare is supported by a complete chain of evidence. This feature empowers physicians not only to make informed decisions regarding diagnoses but also to comprehend the rationale behind them.
Sun Kun, one of the corresponding authors from Xinhua Hospital, stated that the research team is in the process of launching a global AI alliance focused on the diagnosis and treatment of rare diseases. They aim to conduct real-world validation on 20,000 rare disease cases within the next six months, marking a significant step forward in the integration of artificial intelligence into medical practice.
The advancements represented by DeepRare reflect a growing trend towards leveraging AI in healthcare, particularly in specialized areas where traditional diagnostic methods face limitations. As this technology evolves, it holds the potential to greatly enhance diagnostic precision and accessibility, thereby improving patient outcomes on a global scale.
See also
ByteDance Launches Seedance 2.0, Generating High-Quality Videos with Minimal Input
AI Enhances Qualitative Research: Human and LLM Analyses Yield Similar Insights
Agentic AI Enhances Deep Learning Workflows, Automates Hyperparameter Tuning
AI Scientists Surge: Autoscience Institute’s Carl Achieves 75% Conference Acceptance Rate
Swansea University Joins Google DeepMind AI Programme for Disadvantaged Students
