Brain Tumors the Size of a Pinhead—Detected with 97% Accuracy

24 June 2025

Robovision and the Netherlands Cancer Institute’s breakthrough published in Radiology, the world’s top imaging journal

Every year, millions of cancer patients face an unthinkable reality: their disease has quietly spread to the brain before anyone notices. Tiny metastases—some smaller than a grain of rice—are often overlooked by even the best radiologists. But a new AI tool looks to change that.

In a landmark achievement for medical AI, Robovision Healthcare and the Netherlands Cancer Institute (NKI) announce the publication of their pioneering research in Radiology, the flagship journal of the Radiological Society of North America (RSNA). The study unveils how BrainMets AI, a next-generation AI solution capable of detecting regions suspicious for brain metastatic disease with unprecedented precision raises the bar in neuroradiology diagnostics, potentially improving millions of patient outcomes with early detection.

WHY IT MATTERS

• Brain metastases are the most common tumors in the central nervous system, affecting up to 17% of adult cancer patients.

• Early detection is critical: these lesions can grow fast, complicate treatment, and often go unnoticed.

• MRI scans, especially of the brain, generate hundreds of slices per patient.

• Radiologists face overwhelming volumes of MRI and diagnostic complexity, particularly in detecting lesions smaller than 3mm. Even top centers miss small lesions in high-volume cases.

ENTER BRAINMETS.AI

BrainMets.ai, the AI solution developed by Robovision Healthcare in close collaboration with the NKI, achieved 97.4% lesion-level sensitivity — including 93.3% on sub-3mm lesions. In a 260-patient external validation set, the model successfully identified every scan with brain metastases.

Note: Image not to scale

RIGOROUS THREE-PART METHODOLOGY

The research outlined three main components:

A diverse, expertly annotated dataset, including over 1,500 MRI cases from multiple institutions across Europe and the USA.
High-fidelity voxel-level annotations by neuroradiology specialists
A custom model architecture developed in-house by Robovision’s AI team

Like any recipe, the quality of AI comes from the quality of the ingredients and the skill of the chef. With meticulously curated data, accurate annotations and an optimized neural network architecture from Robovision—we were able to create an algorithm that sets a new standard in neuroradiology.”

Stephane Willaert, Head of Robovision Healthcare

REDUCING WORKFLOW BURDENS

Already in clinical validation trials across Europe, and with testing underway in the United States, BrainMets is a concurrent reading aid on track to become commercially available by 2026, pending CE and FDA clearance. It integrates into clinical workflows, offering not only detection but also quantification, segmentation, and longitudinal tracking—saving radiologists time with each case and improving diagnostic confidence.

This tool is a true leap forward. It increases our confidence in detecting even the smallest lesions, supporting more timely and accurate treatment decisions. With automated lesion tracking and volume measurement, we can monitor therapy response more consistently.”

Dr. Laurens Topff, neuroradiologist at the Netherlands Cancer Institute and clinical lead for the project

EARLY DETECTION ADVANTAGES

In a study of 311 scans of 260 patients, BrainMets AI identified all true cases of brain metastases, yielding zero false negatives and fewer than 2 percent false positives.

Research shows that earlier detection of brain metastases can improve survival rates, reduce neurological complications, and expand treatment options.

What happens if lesions are smaller than what can be found with the naked eye?

NEURORADIOLOGY SUPPORT

In addition to detecting very small metastases, BrainMets AI significantly reduces the documentation burden for radiologists by automatically generating clear, structured reports. Lesion data are displayed in an intuitive, clinician-friendly format—complete with snapshots, volumetrics, and growth graphs—supporting faster review, better interdisciplinary communication, and more confident treatment decisions.

We were genuinely impressed by the precision and consistency of BrainMets AI. It has significant potential to help radiologists detect, quantify, and monitor brain metastases with greater confidence — especially in the most challenging or time-pressured cases."

Dr. Michael Iv, Clinical Professor, Neuroradiology, Stanford Medicine

Robovision Healthcare and the Netherlands Cancer Institute (NKI) research publication in Radiology, the journal of the Radiological Society of North America, underscores a critical advance—both for science and for the evolving role of AI in clinical practice.

See the introduction video: here

Introducing BrainMets.ai - YouTube

They’re smaller than a grain of rice—but they can change everything.Brain metastases are among the most deadly—and most overlooked—tumors in cancer care. In ...

VISIT THE BRAINMETS WEBSITE

###

About Robovision Healthcare

BrainMets (brainmets.ai), a product of Robovision's Healthcare division, is creating AI-powered tools for radiologists. With strong partnerships across leading hospitals and a commitment to clinical excellence, BrainMets is redefining the standard of care for neuroradiology.

About Netherlands Cancer Institute (NKI)

The Netherlands Cancer Institute (www.nki.nl) is among the world’s foremost comprehensive cancer centers, combining innovative fundamental, translational, and clinical research with dedicated patient care. In our research institute, around 750 researchers from 45 countries work towards solving the mysteries of health and disease and improving the prospects of cancer patients. We gratefully acknowledge funding from the Dutch Ministry of Health, Welfare and Sport, the Dutch Cancer Society, and individual donors.

Press Contact

Stephane Willaert

Head of Robovision Healthcare at Robovision