Boston Children's Uses OpenAI's AI to Diagnose Rare Diseases

Boston Children's Hospital, one of the largest pediatric medical centers in the US, has deployed an AI system built on OpenAI technology to diagnose rare diseases. Since going live, the system has helped identify more than 40 rare diseases and noticeably reduced the workload of medical staff.

Why rare diseases are the core problem

Rare genetic and hereditary diseases affect a small share of the population, but they are hard to diagnose. A child may see doctors for years and undergo useless treatment until someone notices an unusual pattern of symptoms. In the US, a disease is considered rare if it affects fewer than 200,000 people, which adds up to thousands of different diagnoses.

How the AI assistant works

Boston Children's has integrated a GPT-based system that analyzes a patient's clinical picture: symptoms, test results and the history of the illness. The doctor enters the data, and within seconds the AI compares it against a knowledge base of rare diseases and proposes several diagnostic hypotheses, explaining its reasoning.

The system does not make the decision; the doctor does. But it speeds up the first stage: finding a rare disease among thousands of possibilities.

• Real-time analysis of clinical data
• Comparison of symptoms against a database of rare diseases
• Explanation of why the system produced a particular hypothesis
• Integration with the electronic medical record

The result: 40+ diagnosed cases

For the hospital, this is a historic improvement. More than 40 rare diseases that previously could have gone unrecognized for years are now diagnosed within weeks or months. For each family, that means targeted treatment instead of symptomatic care.

The key point: many of these children live in other parts of the US and came to Boston Children's on referral. The AI has not only improved diagnosis inside the hospital but also made it a destination for the most complex cases.

"We see the result in families' lives.

When we finally find a rare disease, parents get answers and can start the right treatment."

Lighter workload for doctors, better care

Another effect is operational. Previously, a rare-disease physician could spend hours in the literature, consulting specialists at other hospitals and re-taking the patient history. Now the first analysis falls to the AI, which offers 5-10 hypotheses with explanations in minutes. The doctor checks them, refines them and sometimes orders additional tests. The work becomes deeper: not searching for a needle in a haystack, but validating a hypothesis. That is not only faster but also more accurate.

What this means

For rare-disease medicine, this is a prototype of the future: AI as a physician's assistant, not a replacement. The system narrows the search field, and the doctor makes the diagnosis. For US healthcare, this could mean fewer diagnostic delays and fewer unnecessary procedures.