This is an extension of brainchop that converts voxel-based MRI scans to 3D meshes that can be printed. No data is sent to a server. Everything happens in your browser window, on your machine.
- Open the live demo.
- Option 1 The web page automatically loads with a default T1 MRI scan. If you want to use this scan, go to step 5.
- Option 2 If your T1 MRI scan is in NIfTI format, drag and drop the file onto the web page.
- Option 3 If your image is in DICOM format, it may load if you drag and drop the files. If this fails, convert your images with dcm2niix and save the result as a NIfTI format file that brain2print can open.
- Segment your brain scan by choosing a model from the
Segmentation Model
pull-down menu. Not all models work with all graphics cards. TheTissue GWM (High Acc, Low Mem)
is a good starting point. Hopefully, it will accurately segment your brain into gray matter, white matter and cerebral spinal fluid. - Press the
Create Mesh
button and select your preferred settings:
- The
Hollow
option allows you to choose between a solid object or a surface shell (which uses less material to print). - You can choose
Smoothing
to make the surfaces less jagged at the expense of computation time. - You can choose to
Simplify
to reduce the number of triangles and create smaller files.
- Once you have set your preferences, press
Apply
. - You will see the mesh appear and can interactively view it. If you are unhappy with the result, repeat step 6 with different settings. If you want to print the results, press the
Save Mesh
button.
This web application uses some of the latest browser technologies that allow the tissue segmentation model to run on your local graphics card (GPU), regardless of the type of GPU. This is possible via the WebGPU
browser API. Additionally, we leverage ITK-Wasm to turn the tissue segmentation into a 3D mesh. No data ever leaves your machine.
git clone [email protected]:niivue/brain2print.git
cd brain2print
npm install
npm run dev
npm run build
This web page combines three packages developed by our team: