A new study in The New England Journal of Medicine reports that a diagnostic system built on large language models has outperformed experienced clinical geneticists in identifying rare diseases. The finding is significant for a field where patients often spend five to seven years seeking a correct name for their condition.

The research, conducted under controlled conditions, presented both the AI and specialist physicians with identical real patient cases. The system not only identified the correct diagnosis more often but did so more quickly, placing the right answer higher on its list of possibilities.

Clinical geneticists are among medicine's most skilled diagnosticians for these complex cases, making the result a notable marker of progress. The technology analyzes symptoms, genetic data, and clinical notes to navigate a search space of over 7,000 known rare disorders—a task beyond any single doctor's memory.

Researchers emphasize the system is an assistive tool, not a replacement, as real-world medicine involves complexities like incomplete records. Yet the potential impact is substantial. For patients in areas lacking specialist access, such a tool integrated into telehealth or medical records could democratize expert-level diagnostic reasoning.

Companies in genomics and AI are watching closely, as reducing diagnostic delays could lower the enormous costs tied to years of misdiagnosis. Hurdles remain, including concerns about biased training data and establishing clinical trust. However, the evidence is building that AI can help shorten the long, painful search for answers that millions still endure.

Source: Webpronews