About MedSeg
MedSeg is a web application for visualising and segmenting three-dimensional medical images: CT, MRI, and other DICOM volumes. It runs entirely in the browser, with AI inference and custom model training executed on hosted GPUs. It is built and maintained by a small team of radiologists in Oslo, Norway, for other radiologists and researchers. The project's roots go back several years.
Two MedSegs, one team
This site, app.medseg.ai, is the newer multi-user platform: shared projects, hosted AI inference, custom model training, optional PACS integration. The older single-user tool is still available, free and without an account, at medseg.ai. They use the same segmentation engine; the multi-user platform adds collaboration, bigger models, and longer pipelines.
Early access
The multi-user platform is in early access. New accounts are reviewed by a human before they're activated. This is usually fast, but it isn't instant. The point is oversight, not numbers. We want to know who's using the platform, and to flag any use that crosses into territory the tool isn't meant for. When your account is activated you'll get an email at the address you signed up with.
If you'd rather try MedSeg with no waiting list and no account, medseg.ai works in any modern browser today.
What's on the platform
- Pretrained AI models covering CT and MRI: TotalSegmentator, MRSegmentator, FastSurfer, MindGlide, TissUNet, Viola-Unet, BodySegAI, plus a specialty pack for epilepsy work (MELD, AID-HS, nnU-Net FCD, DeepFCD). Happy to add more on request.
- nnInteractive: universal 3-D interactive segmentation, driven by clicks, scribbles, freehand strokes, or bounding boxes.
- Custom nnU-Net training (2-D or 3-D), with optional transfer learning from another model you've already trained in the same project.
- A built-in chat assistant that runs models on demand, summarises series, and answers how-to questions.
- DICOM and NIfTI throughout: you upload either, and you can export masks back as either, including DICOM SEG and PDF reports for PACS routing.
- X-ray and mammography: single-frame DICOMs (chest X-ray, mammography, panoramic dental) load and can be edited, measured, and used for custom 2-D nnU-Net training. The bundled 3-D AI models don't apply to them.
- Phones and tablets: the workspace is responsive and usable on touchscreens, not just desktops.
- VR and AR viewer: open any segmented series in an immersive 3-D viewer powered by WebXR. Works on Meta Quest and similar devices.
Free for research
This cloud platform (app.medseg.ai) is free of charge for personal, academic, and non-commercial research, under the bundled MedSeg Software License.
Commercial and on-premise use
For for-profit clinical services, hosted commercial offerings, or an on-premise install on your own hardware, we offer a separate commercial product. It ships under different terms with a different bundled feature set. Write to us if that's what you need.
Not a medical device
MedSeg is a research tool. It has not been cleared, approved, certified, or registered by the FDA, EMA, MHRA, Health Canada, or any other regulator, and it does not carry a CE mark. Do not use it for any clinical, diagnostic, prognostic, screening, or treatment decision concerning a patient. See the full Terms of Service and Privacy Policy.
How to reach us
The fastest channel is the envelope icon inside the in-app assistant once you're signed in. It opens a short form (Bug / Feature idea / Question / Other) with the option to attach the current conversation as context. From outside the app:
- Email: dlinradiology@gmail.com
- Contact form on medseg.ai
- LinkedIn: linkedin.com/company/medsegmentation
- Twitter / X: @DLinRadiology
Citing MedSeg
If MedSeg helps a paper or project, a reference to medseg.ai is appreciated.