AI Model Helps Clinicians Diagnose Rare Diseases

A new study published in the New England Journal of Medicine reveals that an artificial intelligence model is helping clinicians diagnose patients who have spent years living with unexplained illnesses. Researchers from OpenAI and Boston Children’s Hospital utilized the new tool to review existing genetic data from 18 pediatric patients, successfully identifying conditions that had previously stumped medical professionals.

A Breakthrough for Rare Disease Diagnosis

One participant in the study, 28-year-old Kyra, received a diagnosis of Myofibrillar Myopathy (MFM) after nearly two decades of uncertainty. MFM is a rare group of genetic disorders characterized by progressive muscle weakness. While the condition is currently incurable, Kyra noted that the diagnosis provided her and her family with much-needed clarity and closure.
Researchers hope this technology will eventually assist the millions of Americans impacted by rare diseases. With one in 10 Americans—half of whom are children—affected by a rare condition, the team believes AI can expedite the diagnostic process by sifting through complex genomic information. In this study, the AI tool processed data in approximately six to 10 minutes per case.

The Role of Human Oversight

The study emphasizes that AI is not a replacement for doctors or genetic specialists. The model functions as an additional set of eyes, identifying potential answers that must then be reviewed by experts and confirmed by a certified clinical lab. Catherine Brownstein, a lead researcher and research associate in the division of genetics and genomics at Boston Children’s Hospital, stressed that human guardrails remain essential.
"We're not removing any human guardrails here," Brownstein said. "A human has to review everything that the AI does." Because AI tools can make mistakes or misread information, the researchers maintain that the technology is best used to assist specialists by handling the heavy lifting of data analysis, allowing clinicians to focus their time on specific diagnostic possibilities rather than exhaustive manual searches.

Revisiting Past Genetic Data

The findings suggest that previous genetic test results may be worth re-examining as scientific knowledge evolves. Because the understanding of the genome is expanding rapidly, a test that appeared negative in the past may yield new insights when reviewed with updated data and improved search capabilities.
"A negative genetic test that's negative right now might not be negative in the future," Brownstein explained. While the current study was limited to a small group and retrospective analysis, the research team plans to conduct larger, forward-looking studies across multiple medical centers. For patients like Kyra, the technology offers a promising path forward, though she maintains that the human element of medicine remains vital. "When it comes to health matters that really change your life, you kind of want that human touch present," she said.