IVD-Net: Intervertebral disc localization and segmentation in MRI with a multi-modal UNet

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.

Requirements

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

Running the code

Pre-processing

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 

Some results from our paper

• 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.