Minimal in-browser brain segmentation built only from stock npm packages: @brainchop/mindgrab (models), @niivue/niivue (viewer), @niivue/niimath (meshing) and @niivue/dcm2niix (DICOM). Images never leave the browser.
bun install
bun run dev # live demo at http://localhost:5173
bun run build # static site in dist/Models: MindGrab skull stripping, 16chan18cls, mindmap, mindmap partial-volume (GM/WM/CSF), mindsnap (104 regions). Drop NIfTI images, meshes (STL, GIfTI, …), NiiVue scenes (.nvd), or DICOM files or folders; with several DICOM series a picker suggests the 3D anatomical. Mesh generates a surface over the segmentation (Save writes it as STL); Draw edits labels, Stats reports per-region volume and intensity, Option/Alt-click shows one region alone, and Save writes native or conformed (256³) segmentations. On phones a pane switcher shows one plane at a time.
bun test src # label helper testsEvery push to main builds, tests and pushes dist/ to the gh-pages branch
(.github/workflows/deploy.yml), which GitHub Pages serves at
https://rordenlab.github.io/brainchop-next/ (Settings → Pages → Deploy from a branch → gh-pages).