AI Helps Diagnose 18 Kids After Doctors Were Stumped

For years, doctors at Boston Children's Hospital faced cases that seemed impossible to solve. Eighteen children suffered from illnesses so rare that traditional diagnostic methods couldn't crack the code.

Then OpenAI's o3 model stepped in and changed everything. The AI technology helped identify diagnoses for all 18 patients, including 10 with rare neurodevelopmental diseases and four with neuromuscular disorders.

Catherine Brownstein, a lead researcher from the Manton Center for Orphan Disease Research, called the breakthrough "a total game changer." Her team works with families affected by rare diseases, a challenge facing 30 million Americans.

The process combines human expertise with AI power. Researchers fed the model patient genomes, doctor's notes, symptoms, and possible genetic causes. Humans then reviewed every AI suggestion before making final diagnoses.

Out of 376 undiagnosed cases analyzed, the AI identified nearly five percent of new diagnoses. "Considering how many times these had already been analyzed, that's a huge number, and each one means an answer for a family," Brownstein said.

Kyra Benton lived that reality. She started walking on her tiptoes at age nine, and her ability to run normally disappeared. For over a decade, her health declined without answers.

Just before turning 20, AI helped researchers finally diagnose her with 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

The breakthrough solves a crucial time problem in rare disease research. Suyash Shringarpure, a researcher at OpenAI, explained that a scientist can only dedicate limited hours to each case. Sometimes new research published months later could unlock a diagnosis, but no human can review every patient file against every new study.

AI can process that mountain of information constantly. It never gets tired, never misses a newly published paper, and can spot connections humans might overlook after reviewing the same case multiple times.

This isn't about replacing doctors. The technology serves as a powerful assistant, handling the exhaustive review work while medical professionals make the final calls. OpenAI specifically warns against using the technology for self diagnosis.

For families desperate for answers, though, this collaboration between human expertise and artificial intelligence opens new doors. Every diagnosis means access to proper treatment, understanding of what lies ahead, and finally ending years of uncertainty.