Artificial intelligence may soon help doctors predict whether a common brain tumour is likely to return, using routine tissue slides already available in hospitals, according to a new study that researchers say could make advanced cancer insights accessible to far more patients.

Scientists at the Mayo Clinic and collaborating institutions have shown that AI can analyse standard pathology slides to classify meningiomas—the most common primary brain tumours in adults—and estimate a patient's risk of recurrence. The findings, published in The Lancet Digital Health, suggest that information usually obtained through complex genetic testing could one day be extracted from images already used in routine care.

For thousands of patients diagnosed with meningioma every year, one question often dominates discussions after surgery: will the tumour come back? The answer can influence decisions about follow-up scans, radiation treatment, and long-term monitoring. Yet obtaining detailed biological information about a tumour often requires DNA methylation profiling, an advanced laboratory test that examines chemical markers on DNA to better understand how it behaves. Such testing can be expensive, time-consuming, and unavailable in many healthcare settings.

Using tissue samples, pathology images, and clinical information from 672 patients, researchers trained deep-learning AI models to find hidden patterns in standard haematoxylin and eosin (H&E) slides, which are the pink-and-purple tissue images that pathologists regularly examine under a microscope. Deep learning is a type of AI that learns from large amounts of data and can detect subtle features that may escape the human eye.

"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," says Gelareh Zadeh, M.D., Ph.D., chair of the Department of Neurosurgery at Mayo Clinic in Rochester and the David C. and Flora C. Pratt Distinguished Chief Medical Officer for the Mayo Clinic Platform.

The researchers drew on multiple de-identified datasets, including resources from the Mayo Clinic Platform. Their AI models successfully classified meningioma subtypes and predicted recurrence risk using only routine pathology images. Importantly, these predictions remained useful even after accounting for established risk factors such as tumour grade, patient age, and how completely the tumour had been removed during surgery.

The study also highlighted tumour heterogeneity—differences that exist within different areas of the same tumour. These variations may help explain why some tumours behave aggressively while others remain relatively stable. Understanding these differences is now a key area of cancer research around the world, with earlier studies in breast, lung, and prostate cancers showing that AI can be very helpful in pathology.

Experts say the findings are particularly significant because access to sophisticated molecular testing remains uneven across countries and hospitals. If validated for future prospective studies, AI tools could help bridge that gap by providing advanced tumour insights without requiring specialised genetic laboratories.

Researchers caution that the technology is not yet ready for routine clinical use and requires further testing in real-world settings. Still, the results point toward a future in which personalised cancer care will become available to a broader population, regardless of where patients receive treatment.

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

As healthcare systems continue to search for ways to deliver precision medicine more efficiently, the study offers a glimpse of how AI could transform an ordinary microscope slide into a powerful tool for guiding treatment decisions and improving patient outcomes.