Run pretrained models
Use TotalSegmentator, MRSegmentator, FastSurfer, MindGlide, Viola-Unet, BodySegAI, TissUNet, and specialty tools when the target is already covered.
Browser-based research platform
AI medical image segmentation for CT and MRI
MedSeg combines pretrained segmentation models, interactive 3D editing, text-prompted masks, custom nnU-Net training, and DICOM/NIfTI export in one web workspace.
DICOM + NIfTI
Hosted GPU inference
Not a medical device
Pick the segmentation route that matches the job
Most research workflows need more than one model. MedSeg keeps the common routes in the same project so outputs can be edited, compared, measured, and exported.
Guide nnInteractive
Use clicks, scribbles, freehand strokes, or boxes to create a 3D candidate mask for structures that are not covered by fixed-class models.
Prompt VoxTell with text
Type a target such as "right kidney" or a more specific clinical phrase and get a candidate 3D mask when the VoxTell worker is available.
Train custom nnU-Net
Turn corrected masks into a project-specific 2D or 3D nnU-Net model, then run that model across the rest of the cohort.
What MedSeg is useful for
The strongest fit is research segmentation where you need inspectable masks rather than a black-box score.
1
Cohort annotation
Upload or copy cases, create masks with AI assistance, mark training sources, and export project-level measurements.
2
Model comparison
Run multiple public models on the same imaging series and inspect disagreements directly in the editor.
3
Reusable workflows
Move from first-pass masks to custom model training without leaving the browser or reformatting every file by hand.
Formats and outputs
MedSeg is built around clinical imaging formats, but keeps research exports simple.
| Input | Segmentation output | Export paths |
|---|---|---|
| DICOM CT/MRI, NIfTI, ZIP uploads, and public dataset cases | Multi-label masks that can be edited, locked, measured, merged, and used for training | NIfTI masks, project exports, volume tables, DICOM grayscale/RGB/SEG bundles, and DICOM-wrapped reports where available |
Research use only
MedSeg is not a medical device. It is not cleared, approved, certified, or registered for clinical, diagnostic, prognostic, screening, or treatment decisions. Use it for research and education, and validate every output for your project.