Hetairos is a histology-based artificial intelligence model for predicting central nervous system tumor methylation subtypes
https://www.nature.com/articles/s43018-026-01186-3
Publish Date: 2026-06-10 05:32:00
Source Domain: www.nature.com
- Hetairos is an AI algorithm trained to predict 102 subtypes of Central Nervous System (CNS) tumors using H&E-stained whole-slide images, validated with data from various institutions.
- It was developed using a large dataset of 6,115 slides from 4,961 patients, ensuring compliance with the 2021 WHO classification scheme for CNS tumors.
- The model achieves high accuracy in internal validation (75% for top-1 predictions) and maintains robust performance across ten external cohorts from diverse geographical locations.
- Hetairos’s diagnostic accuracy in external cohorts slightly decreases to 68% for top-1 predictions but remains high for high-confidence cases.
- Compared to neuropathologists, Hetairos shows superior diagnostic accuracy on H&E slides, particularly for common tumor types, although it faces challenges with very rare subtypes.
- The tool assists in narrowing the differential diagnoses, potentially reducing the need for extensive molecular testing and speeding up the diagnostic process, as it takes an average of 12 minutes per slide.
- Hetairos demonstrates prognostic value by stratifying survival rates among different tumor subtypes, showing the importance of granular classification in CNS tumors.
