AI Predicts Brain Tumor Recurrence from Standard Pathology Slides

AI extracts molecular data from routine tissue samples

Researchers at Mayo Clinic have developed artificial intelligence models that can analyze standard pathology slides to classify meningiomas and predict which brain tumors are likely to return after treatment, according to findings published in The Lancet Digital Health.

The deep learning system extracts molecular and prognostic information from hematoxylin and eosin (H&E) slides—the same tissue images pathologists already examine during routine clinical care. This approach could provide insights typically obtained only through DNA methylation profiling, an advanced genetic test that remains costly, time-intensive, and unavailable at many medical centers.

"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, chair of the Department of Neurologic Surgery at Mayo Clinic in Rochester, in a statement first reported by the Mayo Clinic News Network.

Why it matters

Meningiomas are the most common primary brain tumors in adults, but their behavior varies dramatically. Some remain dormant for years while others aggressively recur, making accurate risk assessment critical for treatment planning. Molecular testing can identify high-risk tumors, but access barriers mean many patients make treatment decisions without this information. An AI system that works with existing pathology infrastructure could democratize precision oncology, particularly for hospitals lacking specialized genomic testing capabilities.

Training on 672 patient samples

The research team trained their models using tissue samples, pathology images, and clinical data from 672 patients. The datasets included de-identified information from Mayo Clinic Platform resources. The AI successfully classified meningioma subtypes and predicted recurrence risk using only the standard slides already generated during patient care.

The models identified patterns of tumor heterogeneity—variations within a single tumor—that may explain why some meningiomas behave more aggressively or respond differently to treatment. These AI-based predictions remained valuable even after researchers accounted for traditional clinical factors including tumor grade, surgical resection completeness, and patient age.

Clinical implications for treatment decisions

For patients with meningiomas, recurrence risk directly influences follow-up protocols, imaging schedules, and whether radiation therapy should follow surgery. The ability to assess this risk using existing pathology workflows could accelerate treatment planning and reduce costs.

The researchers emphasized that prospective studies are needed before clinical deployment. However, the findings establish a foundation for more accessible personalized care in meningioma treatment and suggest similar AI approaches could apply to other cancer types.

"The aim is to make these algorithms readily and simply accessible for use globally, improving patient care across many healthcare settings," Zadeh said.

The study details were first reported by the Mayo Clinic News Network on June 8, 2026.