18 children had illnesses so rare doctors were stumped. AI gave them answers
Key Points:
- A study published in NEJM AI highlights how OpenAI’s o3 model helped diagnose 18 children with rare illnesses at Boston Children’s Hospital, including neurodevelopmental and neuromuscular disorders.
- The AI analyzed 376 genomes from undiagnosed patients and identified new diagnoses in nearly 5% of cases, providing answers where traditional methods had failed.
- Researchers used the AI as a diagnostic aid by inputting patient data and symptoms, with final diagnoses confirmed by human experts to ensure accuracy.
- Experts, including lead researcher Catherine Brownstein, called the AI’s application a “total game changer” for diagnosing rare diseases, which affect 30 million people in the U.S.
- Despite skepticism from some patients like Kyra Benton, who was diagnosed with a genetic neuromuscular disorder thanks to the AI, OpenAI emphasizes its technology is not intended for self-diagnosis.