RismadarVoice Reporters, June 18, 2026
Researchers at Boston Children’s Hospital have used artificial intelligence to identify the causes of rare diseases in 18 children whose conditions had previously defied diagnosis.
The breakthrough, published in NEJM AI, a journal focused on artificial intelligence in medicine, showed that OpenAI’s o3 model helped researchers uncover genetic explanations for illnesses that had remained unresolved despite years of investigation.
The study involved 376 patients with rare diseases who had undergone extensive genetic analysis without receiving definitive diagnoses. Researchers combined clinical notes, symptom descriptions and selected genetic data, which were then analysed by the AI model. Medical experts reviewed the results before confirming any diagnosis.
According to the research team, the process led to new diagnoses in 18 cases, including patients with rare neurodevelopmental disorders, neuromuscular diseases, early childhood psychosis and unexplained sudden deaths.
Catherine Brownstein, scientific director of genetic investigations at the Manton Centre for Orphan Disease Research at Boston Children’s Hospital, described the findings as a significant advancement in rare disease diagnosis.
She noted that while the diagnostic rate represented about five per cent of the cases studied, the achievement was remarkable because the patients’ genetic data had already been reviewed multiple times by specialists.
The Manton Centre supports more than 3,500 people affected by rare diseases across the United States and internationally. Researchers routinely re-examine patient genomes as new scientific discoveries emerge, but identifying disease-causing genetic mutations remains a complex and time-consuming process.
One beneficiary of the research was 20-year-old Kyra Benton, who had spent years searching for answers after developing symptoms that affected her mobility and heart function. Researchers eventually identified her condition as myofibrillar myopathy, a rare genetic disorder that causes progressive muscle deterioration.
Benton said she had largely accepted that she might never receive a diagnosis before researchers contacted her with the findings.
Medical experts not involved in the study described the results as promising. Adam Rodman, a physician and AI researcher at Beth Israel Deaconess Medical Centre, said the technology could help reduce backlogs of unresolved cases and accelerate genetic investigations.
Researchers, however, stressed that AI is not a substitute for medical professionals. They said all findings require thorough human review and that the technology should be used as a tool to support clinical decision-making rather than independently diagnose patients.
The study also revealed that seven of the identified diagnoses had previously been made by individual treatment teams but were not widely shared within the medical community, highlighting the importance of improved data access and collaboration.
OpenAI’s health team said the findings demonstrate how commercially available AI systems can assist healthcare professionals in navigating complex medical information and uncovering diagnoses that might otherwise remain undetected.
The researchers cautioned that receiving a diagnosis is often only the first step in managing a rare disease, but said the information can help families better understand a condition and potentially access future treatments as they become available.
