Researchers have unveiled an artificial intelligence model that classifies more than 100 brain tumor subtypes in just 12 minutes, drastically reducing the standard 12-day wait for molecular testing.
Named Hetairos, the system analyzes routine hematoxylin and eosin pathology slides to predict DNA methylation patterns. Developers trained the model on over 11,000 digital slides from nearly 10,000 patients across four continents. In high-confidence scenarios, Hetairos delivers accurate classifications for up to 70% of cases, providing rapid triage data where advanced lab infrastructure is limited or delayed.
During clinical validation, the AI achieved a diagnostic accuracy of 0.68, more than doubling the 0.30 accuracy of board-certified neuropathologists working without molecular data. While DNA methylation testing remains the clinical gold standard, Hetairos is engineered to support expert review, narrow differential diagnoses, and accelerate treatment planning. The breakthrough demonstrates that machine learning can extract molecular-level insights from standard tissue samples, signaling a major efficiency shift in neuro-oncology.