AI in medicine helps Boston Children's Hospital crack 18 rare disease diagnosis cases

AI in medicine delivers breakthrough for families searching for answers

Researchers at Boston Children's Hospital and Harvard have achieved a significant milestone in rare disease diagnosis by using OpenAI's o3 Deep Research model to identify 18 new diagnoses among children whose illnesses had remained mysteries for years 1

1

. The findings, published in NEJM AI study, demonstrate how AI-driven medical advancement can help clinicians navigate the complexity of genetic data and scattered medical literature to solve cases that had undergone multiple rounds of expert review without success 2

2

.

The Manton Center for Orphan Disease Research analyzed 376 de-identified pediatric cases of patients who lacked diagnoses despite previous genetic testing and clinical evaluation 1

1

. Catherine Brownstein, the scientific director of the genetic investigations arm at the center, called the results "a total game changer," noting that the OpenAI o3 model helped achieve approximately 5 percent new diagnoses—a substantial figure considering these unsolved genetic cases had already been analyzed multiple times 3

3

.

How AI tackles the complexity of undiagnosed genomes

Finding the genetic cause of a disease involves navigating roughly 20,000 protein-coding genes in the human genome. While sequencing a patient's full genome has become straightforward, matching errors or abnormalities in specific genes with illnesses remains extraordinarily complex 1

1

. Suyash Shringarpure, a technical researcher at OpenAI focusing on health applications, explained that cases may remain unsolved initially, but a year later a published paper might clarify the link between a gene and disease—information that researchers with limited time per case might miss.

The research team provided the o3 system with clinicians' notes about each case, descriptions of patient symptoms, and a filtered list of genes that might be responsible for the symptoms. The model synthesized this information with medical literature to identify potential diagnoses, which the human research team then reviewed for final verification 2

2

. This approach demonstrates how AI serves to augment rather than replace clinical expertise, with human verification remaining essential to the diagnostic process.

Real impact: From diagnostic odyssey to answers

Of the 376 cases spanning four disease areas, the team identified new diagnoses for 10 patients with neurodevelopmental disorders, four patients with neuromuscular diseases, two children who had died suddenly without further specification, and two patients with early childhood psychosis illnesses 1

1

. One patient, Kyra Benton, spent years navigating a diagnostic odyssey after her mother noticed she moved differently from her peers at age nine. After visits to specialists in New York City and Boston Children's Hospital yielded no answers, Benton faced severe heart problems and underwent a tracheotomy at 13. She had come to terms with never knowing her diagnosis until researchers called last summer, just before her 20th birthday, to reveal that the AI analysis had identified myofibrillar myopathy—a progressive genetic neuromuscular disorder causing muscle fibers to break down 1

1

.

Broader implications for precision medicine and rare diseases

The NEJM AI publication represents part of a broader initiative at Boston Children's Hospital, which announced in May that its AI efforts had led to over 40 rare disease diagnoses previously deemed unsolvable 2

2

. OpenAI has committed $50 million to support the hospital's AI initiatives, which began in early 2025. The hospital reports that more than one-third of its employees now use AI tools in their daily work, resulting in approximately 60,000 hours saved, valued at over $7 million 2

2

.

John Brownstein, Chief Innovation Officer at Boston Children's Hospital, emphasized that the challenge in diagnosing complex cases often stems from cognitive limits. "We combine genetic information, phenotypic information, literature search, and the reasoning of AI to diagnose rare genetic diseases for families that were once left without any answers," he stated 2

2

. With rare diseases affecting an estimated 300 million people globally, and many families navigating lengthy diagnostic processes, this approach to precision medicine offers new hope. The Manton Center works with over 3,500 individuals across all 50 states and around the world affected by rare diseases, routinely screening their genomes against newly identified genes 1

1

. As AI models continue to improve and more medical literature becomes available, the potential to revisit previously unsolved cases grows, suggesting that families who once faced dead ends may find answers in the future.