Artificial intelligence tools demonstrate great potential to transform healthcare when applied correctly. Although virtual assistants on the internet sometimes offer dubious medical information, the technology proves effective where human capacity reaches its limits.
New research conducted by professionals at Boston Children’s Hospital, in the United States, proves this ability by revealing that AI systems diagnosed 18 children with previously unknown health conditions. Thanks to these innovations, patients with unusual diseases and manifestations are able to obtain a medical opinion significantly faster.
“This represents a significant milestone,” Catherine Brownstein, the study’s lead researcher, told NBC. She emphasized that “given the frequency with which these conditions were previously investigated, this is a significant volume of diagnoses, and each one offers a resolution for the families involved.”
The initiative is highly valuable, as highlighted by the researcher, as the analysis of an extensive volume of genomic information is comparable to the difficult search for something tiny in a huge place.
The study, published in the journal “NEJM AI”, resulted from a collaboration between the hospital’s Center for Rare Diseases and OpenAI, and demonstrated the ability of technologies developed by the company to detect anomalies in the genomes of the patients examined.
Using OpenAI’s “o3 Deep Research” model, scientists were able to examine hundreds of genomic sequences from individuals who remained without a clear diagnosis for rare diseases, resulting in almost 5% of new findings. This speed represents a considerable relief for families, who often face long and exhausting years searching for a diagnosis for rare conditions, where time is crucial.
The investigation of the genetic origin of a disease in thousands of patient data, when carried out by humans, can take several days, given the existence of approximately 20 thousand genes that encode proteins in the human genome. The team submitted the genomes of 376 undiagnosed patients to the o3 system, enriching the analysis with clinical notes, detailed descriptions of symptoms and a selection of potential genes.
Among the results, new diagnoses were established for ten patients with rare neurodevelopmental illnesses, four with neuromuscular disorders and two with psychosis that manifested early in childhood. Furthermore, the causes of the illnesses of two children who died abruptly, without prior explanation, were also determined.