AI Model Reviewed Genetic Data of 18 Patients, Yielding Few New Diagnoses in Small Study

Researchers from OpenAI and Boston Children's Hospital applied a newly developed AI model to the existing genetic data of 18 pediatric patients, many now adults, and identified diagnoses that had eluded clinicians for years. The study was published Thursday in the New England Journal of Medicine.

One in 10 Americans, more than 30 million people, half of whom are children, have a rare disease, the researchers noted.

For 28-year-old Kyra, the model produced an ultra-rare diagnosis of Myofibrillar Myopathy after nearly two decades of uncertainty. Myofibrillar Myopathy is a group of genetic disorders that causes progressive muscle weakness. Only 60 patient cases of the condition have ever been published, according to CureMFM13.

"It felt very surreal at the time because I just didn't expect to get an answer in my lifetime, and I think my family didn't expect it either," Kyra told ABC News. The AI model processed each case in about six to 10 minutes. Experts made the final diagnosis in every instance, and each diagnosis was confirmed by a certified clinical lab before families were informed.

Catherine Brownstein, a lead researcher and research associate in the division of genetics and genomics at Boston Children's Hospital, said the tool functions as an additional set of eyes rather than a replacement for clinicians. "A human has to review everything that the AI does," Brownstein told ABC News.

"We can apply our human time to more specific things, like reviewing the data, rather than going down rabbit holes chasing things that might be possibilities for a diagnosis," she added.

The study was a retrospective review of existing cases and did not test the tool in real time. The number of new diagnoses produced was small, and the study did not measure whether the AI tool saved time, lowered costs, or changed patients' care. Brownstein noted that genetic knowledge advances rapidly, making older negative results worth revisiting.