Repository containing the source code of the IVD-Net segmentation network (link) that we proposed for the MICCAI 2018 IVD segmentation challenge. This architecture was used to segment the inetervertebral disc in multi-modal MRI images. Nevertheless, it can be used to segment any other structure/s in a multi-modal image setting.

• The code has been written in Python (3.5) and pyTorch.
• You should also have installed torchvision and scipy

If you just want to run the IVD dataset with this code, you simply need to convert the 3D volumes to 2D slices/images. Then, the structure to save the images should be (note that val and test folders will have the same structure than train):

--| MainFolder
----| train
--------| Fat/
------------| ImgName_xxxx0.png
------------| ImgName_xxxx1.png
------------| ....
--------| GT/
------------| ImgName_xxxx0.png
------------| ImgName_xxxx1.png
------------| ....
--------| Inn/
------------| ImgName_xxxx0.png
------------| ImgName_xxxx1.png
------------| ....
--------| Opp/
------------| ImgName_xxxx0.png
------------| ImgName_xxxx1.png
------------| ....
--------| Wat/
------------| ImgName_xxxx0.png
------------| ImgName_xxxx1.png
------------| ....
----| Val/
----| Test/

Then you simply have to write in the command line:

python main.py 

• Dolz J, Desrosiers C, Ben Ayed I. IVD-Net: Intervertebral disc localization and segmentation in MRI with a multi-modal UNet. arXiv preprint arXiv:1811.08305. 2018 Nov 19.