In a study on diagnosing rare diseases, DeepRare surpasses the performance of doctors.

      DeepRare, an advanced AI system that incorporates 40 specialized tools, surpassed medical experts in recognizing rare diseases in a direct comparison detailed in a Nature publication.

      For millions afflicted by rare ailments, obtaining a diagnosis is a complex journey. Patients often circulate between general practitioners and specialists for years, sometimes spanning decades, as they attempt to connect symptoms that don't fit conventional descriptions.

      Eighty percent of rare diseases are genetically based, but many remain undiagnosed until significant biological harm has occurred. The challenge lies not in the scarcity of data, but rather in locating the rare insights among vast medical information.

      A recent study released in Nature indicates that artificial intelligence may facilitate this search. Researchers from the School of Artificial Intelligence at Shanghai Jiao Tong University and Xinhua Hospital crafted DeepRare, an AI designed to emulate the way human clinicians navigate diagnostic uncertainties.

      In a direct matchup with five seasoned physicians, each with over ten years of experience, the AI demonstrated superior accuracy overall.

      The statistics are impressive: DeepRare accurately diagnosed conditions in its first attempt 64.4% of the time, in contrast to the 54.6% success rate of the doctors. When given three potential diagnoses instead of just one, the AI's diagnostic success rose to 79%, compared to 66% for the human experts.

      Importantly, the physicians agreed with the AI's reasoning 95.4% of the time, indicating that the system not only arrives at correct answers but does so in a manner that resonates with clinical professionals and is seen as medically valid.

      What sets DeepRare apart from previous diagnostic AI systems is its design. Instead of utilizing a black-box classification model, it incorporates 40 specialized digital tools and adheres to a clearly defined reasoning process.

      The system formulates diagnostic hypotheses, evaluates them against patient data, searches extensive medical literature, analyses genetic variations, and continuously updates its conclusions before prioritizing options.

      This procedure reflects the cognitive processes of human diagnosticians, yet benefits from comprehensive access to medical knowledge and a computational speed unattainable by humans.

      DeepRare has already progressed beyond the research phase. Since July 2025, it has been implemented on an online diagnostic platform, with over 600 medical institutions globally registered to use it.

      The research team aims to further validate the system with 20,000 real-world cases and establish a global alliance focused on rare disease diagnostics. Importantly, the authors stress that the system is meant to enhance clinicians’ work rather than replace them, highlighting both the limitations of AI technology and the essential human factor in medicine.

      The implications for patients are significant. Around 300 million individuals worldwide suffer from rare diseases, and the average diagnostic journey can take five years or longer.

      Each year of delay translates to a year of uncertainty, incorrect treatments, and worsening organ damage. An AI system capable of reducing weeks or months from this timeline, and revealing potential diagnoses that may otherwise be missed, could transform the initial experience of managing a rare condition.