Free brain tumor MRI classifier
Drop in a brain MRI image and a small neural network flags the most likely pattern — glioma, meningioma, pituitary tumor, or no tumor — with a confidence score. It runs entirely in your browser: the image never leaves your device. Informational only, not a diagnosis.
Drop a brain MRI image here, or
JPG or PNG of a single brain MRI slice (axial works best). Processed on your device — nothing is uploaded.
No MRI handy? Try a sample:
Sample images from the Brain Tumor MRI Dataset by Masoud Nickparvar — CC BY 4.0.
Informational only — not a medical diagnosis. This tool classifies image patterns and can be wrong. Always consult a qualified radiologist or physician.
Opens: A single brain MRI slice as a JPG or PNG (axial views work best). · Files are processed in your browser — nothing is uploaded.
What this tool does
Four classes, with the confidence for all four
Glioma, meningioma, pituitary tumor or no tumor — and not just the winner. The full confidence breakdown across the four categories is shown, because a 40/35/15/10 split and a 97/1/1/1 split mean very different things and a single label hides which one you got.
The model runs inside your browser
A MobileNetV2 network exported to ONNX and executed by ONNX Runtime Web through WebAssembly. No GPU required, no server call, and the image never leaves your device — the page loads the model, not your file.
Small enough to be quick
The model file is 8.9 MB, so it downloads once and classifies in about a second on an ordinary laptop. It works on a phone, and once the page has loaded it keeps working without a connection.
A measured accuracy figure, with its caveat
95.3% on a held-out test set of 1,600 brain MRI images, 400 per class, that the model never saw during training on 5,600 labelled images. That number comes from one curated public dataset; a real scan from a different scanner, sequence, plane or quality will do worse, and the output is a hint rather than a finding.
What to feed it
One brain MRI slice as a clear JPG or PNG. Axial views resembling standard screening images are what it was trained on; a photograph of a screen, a sagittal knee or a CT is outside its world — and it will still return a confident-looking answer.
What it does not do
It classifies a single slice. It does not locate a tumor, measure it, grade or stage it, read a whole series, or consider your history — and it has only four options, so anything that is not one of them gets forced into one of them. It is an informational pattern-matching demo, it can be wrong, and it is not a diagnosis: only a qualified radiologist or physician can make one.
How to use it
- 1.Drop in a JPG or PNG of one brain MRI slice, or click to choose a file.
- 2.The model runs locally in your browser (ONNX / WebAssembly) — nothing is uploaded.
- 3.Read the most likely class and the confidence breakdown across all four categories.
- 4.For your real scan: create a free account and get a plain-language AI report you can ask questions about.
FAQ
- What can this brain tumor classifier detect?
- It sorts a brain MRI slice into four categories from its training data: glioma, meningioma, pituitary tumor, or no tumor. It does not measure size, stage or location.
- Is my MRI image uploaded anywhere?
- No. The model runs in your browser with WebAssembly, so the image is processed on your device and never sent to a server.
- Is this a diagnosis?
- No. It is an informational pattern-matching tool that can be wrong. Only a qualified radiologist or physician can diagnose a brain tumor. Always seek professional medical advice.
- How accurate is it?
- It gets 95.3% right on its held-out test set — 1,600 brain MRI images (400 per class) the model never saw while training. It is a MobileNetV2 network fine-tuned on 5,600 labelled images and kept small (8.9 MB) so it can run in your browser. That score comes from one curated public dataset, so expect less on your own scan: real MRIs vary by scanner, sequence, plane and quality. Treat the result as a rough hint, not a clinical finding.
- What image should I use?
- A single brain MRI slice as a clear JPG or PNG. Axial (top-down) views similar to standard MRI screening work best.
Want your scan explained in plain language?
The viewer is free — and so is understanding your radiologist’s report. Create an account, upload your scan (de-identified in your browser) and your report, and every finding is mapped onto the images, explained in plain language and scored — no card needed. An independent AI reading of the images is a paid extra. Informational only — not a medical diagnosis.