DeepRare AI Beats Doctors in Rare Disease Diagnosis
Key Points
- DeepRare integrates 40 specialized tools and follows a reasoned diagnostic workflow.
- In a study, the AI identified rare diseases on its first suggestion 64.4% of the time versus 54.6% for doctors.
- When offering three suggestions, DeepRare achieved 79% accuracy compared with 66% for clinicians.
- Physicians endorsed the AI’s reasoning 95.4% of the time.
- Since July 2025, more than 600 medical institutions worldwide have accessed the platform.
- The system aims to augment clinicians, not replace them, serving as a decision‑support tool.
- Rare diseases affect roughly 300 million people, with diagnostic journeys often lasting five years or more.
DeepRare, an AI system that combines 40 specialized tools and a reasoned workflow, outperformed seasoned physicians in a head‑to‑head study of rare disease diagnosis. The system identified the correct disease on its first suggestion 64.4% of the time versus 54.6% for doctors, and reached 79% accuracy when offering three suggestions compared with 66% for clinicians. Physicians endorsed the AI’s reasoning 95.4% of the time. Since its July 2025 launch, more than 600 medical institutions have accessed the platform, which aims to augment—not replace—human diagnosticians and could shorten the years‑long diagnostic odyssey for millions worldwide.
Innovative AI Approach to Rare Disease Diagnosis
Researchers at Shanghai Jiao Tong University’s School of Artificial Intelligence and Xinhua Hospital created DeepRare, an artificial‑intelligence system designed to mimic the diagnostic reasoning of human doctors. Unlike earlier black‑box models, DeepRare integrates forty specialized digital tools and follows an explicit workflow that forms hypotheses, tests them against patient data, searches medical literature, analyzes genetic variants, and iteratively revises conclusions before ranking possibilities.
Head‑to‑Head Performance Against Experienced Physicians
In a direct comparison with five physicians each having more than a decade of practice, DeepRare demonstrated higher diagnostic accuracy. The AI correctly identified the disease on its first suggestion 64.4% of the time, while doctors achieved 54.6% accuracy. When allowed to present three possible diagnoses, DeepRare’s success rose to 79% versus 66% for the clinicians. Moreover, physicians endorsed the AI’s reasoning 95.4% of the time, indicating that the system’s explanations were persuasive and medically sound.
Real‑World Deployment and Reach
Since July 2025, DeepRare has been available on an online diagnostic platform. More than six hundred medical institutions worldwide have registered to use the system, moving it beyond the laboratory into everyday clinical practice. The research team plans further validation with twenty thousand real‑world cases and intends to launch a global rare‑disease diagnostic alliance.
Potential Impact on Patients
Rare diseases affect roughly three hundred million people worldwide, and patients often endure a diagnostic odyssey that can stretch five years or longer. Each year of delay can mean missed treatment opportunities and accumulating organ damage. By offering rapid, accurate suggestions and surfacing possibilities that might be overlooked, DeepRare has the potential to trim weeks or months from that timeline, improving outcomes for millions.
Role of AI in Clinical Care
The authors stress that DeepRare is not intended to replace clinicians but to augment diagnostic workflows. This stance acknowledges both the technical limits of AI and the essential human elements of medicine, positioning the system as a decision‑support tool that enhances, rather than supplants, physician expertise.