Hetairos is a histology-based artificial intelligence model for predicting central nervous system tumor methylation...

Area of Science:

• Neuropathology
• Artificial Intelligence
• Oncology

Background:

• Molecular testing, particularly methylation profiling, offers high diagnostic accuracy for central nervous system (CNS) tumors but is resource-intensive.
• Conventional hematoxylin and eosin (H&E) histopathology is globally accessible but lacks the granularity of molecular methods.

Purpose of the Study:

• To develop and validate an artificial intelligence (AI) algorithm, Hetairos, capable of predicting 102 methylation-based CNS tumor subtypes directly from digital H&E slides.
• To assess the performance of Hetairos compared to neuropathologists and evaluate its utility in routine diagnostics.

Main Methods:

• Development and validation of the Hetairos AI algorithm using a dataset of 9,606 patients and over 11,000 digital H&E slides from 11 international centers.
• Comparative analysis of Hetairos's diagnostic accuracy against five board-certified neuropathologists using histology-only assessment.
• Prospective evaluation of Hetairos in a routine diagnostic setting to determine its impact on turnaround time.

Main Results:

• Hetairos accurately predicted 102 CNS tumor subtypes from H&E slides with an overall accuracy of 0.87 for its highest-rated predictions, identifying 50-70% of cases with high confidence.
• The AI algorithm demonstrated superior performance compared to neuropathologists (0.68 vs. 0.30 accuracy).
• Prospective evaluation showed a significant reduction in diagnostic turnaround time from 12 days (molecular testing) to 12 minutes.

Conclusions:

• Hetairos provides a rapid and accurate method for classifying CNS tumors using widely available H&E histopathology, bridging the gap between traditional methods and molecular profiling.
• The AI tool effectively supports diagnostic decision-making for both pediatric and adult CNS tumors by refining differential diagnoses and guiding molecular testing strategies.
• Hetairos has the potential to enhance efficiency and accessibility in CNS tumor diagnostics globally.