AI Helps Diagnose 18 Kids With Rare Diseases in Boston

For years, Kyra Benton walked on her tiptoes and struggled to run normally, while doctors searched desperately for answers. Just before her 20th birthday, artificial intelligence helped solve the mystery that had stumped medical experts since she was nine years old.

A groundbreaking study published this week in NEJM AI reveals that OpenAI's o3 model helped diagnose 18 children at Boston Children's Hospital whose rare conditions had left doctors stumped. The cases included 10 children with rare neurodevelopmental diseases, four with neuromuscular disorders, and others whose symptoms had remained medical mysteries for years.

The research team analyzed 376 genomes from patients with undiagnosed rare diseases. The AI identified nearly five percent of new diagnoses, cases that had already been reviewed multiple times by human experts without success.

"Considering how many times these had already been analyzed, that's a huge number, and each one means an answer for a family," said Catherine Brownstein, lead researcher from the Manton Center for Orphan Disease Research at Boston Children's Hospital. She called the technology "a total game changer."

The challenge with rare diseases is time. New research constantly emerges linking specific genes to specific conditions, but researchers can only dedicate limited hours to individual cases. A patient's genome analyzed last year might finally make sense today based on a study published last month.

That's where AI steps in. The model reviewed patient genomes against newly identified genes, cross-referencing doctors' notes, symptoms, and potential genetic causes. Human experts then verified every diagnosis the AI suggested.

Benton finally learned she has myofibrillar myopathy, a progressive genetic neuromuscular disorder. "Quite frankly, I'm the type of person that's not all that much in favor of AI," she admitted. "On the other hand, I do acknowledge that it does have its advantages."

The Ripple Effect

This breakthrough extends far beyond 18 families. Thirty million Americans live with rare diseases, many spending years searching for answers while their health declines. The AI model doesn't replace doctors but acts as a tireless research assistant, continuously scanning new medical literature and genetic discoveries that human experts might miss.

The technology works best as a collaboration. Doctors provide the clinical insight, patient history, and final judgment, while AI handles the exhaustive task of connecting dots across thousands of research papers and genetic databases.

As genetic research accelerates and more disease-causing genes get identified, AI could help thousands of families escape the painful limbo of not knowing what's wrong with their child. Each diagnosis means access to proper treatment, support groups, and most importantly, hope.