A groundbreaking artificial intelligence system developed jointly by researchers at MIT and Johns Hopkins University has achieved a 98% accuracy rate in diagnosing more than 3,000 rare diseases — a feat that typically takes human specialists years to accomplish.
The system, called MedInsight, was trained on over 50 million anonymized patient records and uses a novel multi-modal learning approach that combines genomic data, medical imaging, and clinical notes.
What makes MedInsight different:
In clinical trials spanning 12 hospitals across the United States and Europe, MedInsight correctly identified rare conditions in an average of 4.2 hours — compared to the current average diagnostic odyssey of 4.8 years for rare disease patients.
Dr. Elena Rodriguez, lead researcher at MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL), said the system was particularly effective at identifying patterns across thousands of data points simultaneously — something human doctors simply cannot do at scale.
The technology is expected to receive FDA fast-track review and could be available in select medical centers as early as 2027. Researchers estimate the system could help the estimated 300 million people worldwide living with rare diseases.