Chinese research team develops AI medical diagnosis system for rare diseases with solid evidence

On the early morning of February 19, Professor Sun Kun and Professor Yu Yongguo from Shanghai Jiao Tong University School of Medicine Affiliated Xinhua Hospital, along with Professor Zhang Ya and Associate Professor Xie Weidi from Shanghai Jiao Tong University, jointly developed an artificial intelligence system capable of diagnosing rare diseases. The related results were published in the top international journal Nature on February 19 Beijing time. This system, called DeepRare, is regarded as the world’s first “traceable reasoning process” intelligent doctor for rare diseases.

For a long time, rare diseases have been difficult to diagnose, and patients often visit multiple hospitals without a clear diagnosis. Previous medical AI could provide diagnostic results, but it was like a doctor who only gives conclusions without explanations, making doctors hesitant to trust it. The biggest difference of this new system is that each diagnosis it makes is accompanied by a complete “evidence chain”—similar to an experienced doctor explaining step-by-step why a certain diagnosis is made during ward rounds.

The system’s “brain” is filled with top-tier medical knowledge and real case data from around the world. It does not simply look up information but thinks like a human doctor: proposing hypotheses, searching for evidence to verify, self-correcting if something is wrong, and repeatedly analyzing before reaching a conclusion.

Test data shows that even without genetic testing results, relying solely on the patient’s clinical symptoms, this system can make judgments with a first-diagnosis accuracy of 57.18%, nearly 24 percentage points higher than the best international model previously available. This means that in primary hospitals lacking genetic testing equipment, doctors can use it for preliminary screening of rare disease patients. When combined with genetic data, the system’s diagnostic accuracy can exceed 70%.

This system is not just theoretical; it has been online since July last year as a diagnostic platform. In half a year, over 600 medical institutions worldwide have registered to use it, from major hospitals in China to top laboratories in Europe and America. At Xinhua Hospital in Shanghai, it has already been deployed internally and will soon be officially operational to assist doctors in identifying gaps in diagnosis and prevent missed diagnoses in rare disease patients.

Professor Sun Kun stated that the development team plans to initiate a global rare disease diagnosis and treatment alliance, and within the next six months, further validate the system with 20,000 real cases to help more patients.

(Source: Jiemian News)