Researchers at Mayo Clinic have developed an artificial intelligence (AI) model that can analyze standard hematoxylin and eosin (H&E) pathology slides to classify meningiomas and predict a patient's risk of tumor recurrence. The study, published in The Lancet Digital Health, suggests AI may be able to extract molecular and prognostic information that currently requires DNA methylation testing.
Using pathology images, tissue samples, and clinical data from 672 patients, the researchers trained deep-learning models to identify tumor subtypes and recurrence risk from routine slides already used in clinical practice. The AI predictions remained informative even when accounting for traditional factors such as tumor grade, extent of surgical removal, and patient age.
Researchers also found the models could identify patterns of tumor heterogeneity—differences within the same tumor—that may help explain why some meningiomas behave more aggressively than others.
"This is one of the many studies where we can harness the strength of digital pathology by capturing the last two decades of genomic and molecular knowledge into AI algorithms," said Gelareh Zadeh, MD, PhD., chair of the Department of Neurologic Surgery at Mayo Clinic in Rochester and the David C. and Flora C. Pratt Distinguished chief medical officer for Mayo Clinic Platform.
AI Extracts Molecular Insights from Routine Pathology Slides
For pathology laboratories, the findings highlight the growing potential of digital pathology and AI to deliver molecular-level insights without requiring specialized genomic testing. DNA methylation profiling can provide valuable diagnostic and prognostic information but remains unavailable in many healthcare settings due to cost, infrastructure, and expertise requirements. An AI-based approach that leverages existing pathology workflows could help make advanced tumor characterization more broadly accessible.
The study adds to a growing body of evidence that AI may be able to unlock clinically useful information from routine pathology slides, potentially supporting treatment planning, follow-up strategies, and decisions about adjuvant therapies such as radiation.
While the technology requires additional prospective validation before clinical adoption, researchers say the work lays the groundwork for future AI tools that could help pathologists and clinicians obtain deeper biological insights from standard tissue specimens while maintaining physician oversight.
This article was created with the assistance of Generative AI and has undergone editorial review before publishing.
