DeepRare AI outperforms doctors on rare disease diagnosis in head-to-head test
Rare diseases are complex medical disorders that are notoriously difficult to diagnose because many present with a wide variety of symptoms that can overlap with more common illnesses. Currently, around 300 million people globally are affected by these disorders, and diagnosis often takes five years or more.
During that time, many patients will endure a diagnostic journey marked by repeated referrals, misdiagnosis, and unnecessary medical interventions. But now researchers have developed an AI system called DeepRare that identifies these diseases more accurately than experienced doctors.
Digital experts
To build DeepRare, the team moved away from the way traditional AI works, with one program trying to solve a problem on its own, as they detail in a paper published in the journal Nature. Instead, they designed a system that uses 40 different specialized digital tools that can analyze everything from a patient’s DNA to official medical databases and a doctor’s handwritten notes.
At the heart of the system is a central AI host that ensures all the different tools work together to find the right answer.
DeepRare’s first test was conducted on 6,401 clinical cases in which the diagnosis was already known. By feeding the AI the same symptoms and DNA data that original doctors had years ago, the researchers found that DeepRare could have correctly identified the right disease much earlier in the process. It also outperformed 15 other existing diagnostic tools.
Surpassing the experts
But the big test came when the researchers conducted a side-by-side evaluation using a smaller group of 163 difficult cases. Five experienced doctors, each with more than a decade of practice, were given the same information as DeepRare. The AI correctly identified the disease in its first try 64.4% of the time, while the doctors achieved the same accuracy 54.6% of the time.
“DeepRare is one of the first computational models to surpass the diagnostic performance of expert physicians in the complex task of rare-disease phenotyping and diagnosis,” commented the team in their paper.
Even when DeepRare did not get the correct answer the first time around, it was still close, achieving a high Recall@3 score, which means that the correct diagnosis was almost always within its top three suggestions. Ten rare disease specialists were asked to look at the AI’s step-by-step reasoning, and they agreed with its logic 95.4% of the time.
“Our work not only advances rare disease diagnosis but also demonstrates how the latest powerful large-language-model-driven agentic systems can reshape current clinical workflows.”
New App for Traumatic Brain Injury (TBI) Evaluation
You can access the app here:
https://apps.apple.com/app/id6753718476
