AI tool predicts meningioma recurrence risk from routine pathology slides

Researchers at the Mayo Clinic have developed deep‑learning models that analyze standard haematoxylin‑and‑eosin (H&E) stained tissue slides to classify meningioma subtypes and estimate the likelihood of tumor recurrence. The study, involving 672 patients and published in The Lancet Digital Health, shows that the AI system can extract molecular and prognostic information that normally requires costly DNA‑methylation profiling. By using images already generated in routine clinical practice, the approach could help clinicians decide on post‑surgical therapies such as radiotherapy without needing specialized genetic tests. The findings suggest broader potential for digital pathology to make advanced tumor diagnostics more accessible.

The AI models were trained on de‑identified datasets from the Mayo Clinic platform and validated across multiple independent cohorts. Lead investigator Gelareh Zadeh, MD, PhD, highlighted that further validation could enable the tool to assist decision‑making for patients worldwide, reducing reliance on expensive molecular assays.