Six-month infant brain tissue segmentation using three dimensional fully convolutional neural networks and pseudo-labelling
Implementation of the nic_vicorob team for addressing the MICCAI Grand Challenge on 6-month infant brain MRI segmentation iSeg2017.
Once the repository has been clone/downloaded, there will be only the log
and models
folders and the iSeg2017.ipynb file. Add the folders datasets
(folder containing the testing and training sets provided by the challenge organisers), results
and refined-results
. The resulting tree should look as indicated below.
.
├── datasets
│ └── iSeg2017
│ ├── iSeg-2017-Testing
│ └── iSeg-2017-Training
├── iSeg2017.ipynb
├── log
│ └── iSeg2017
├── models
│ └── iSeg2017
├── refined-results
│ └── iSeg2017
│ └── iSeg-2017-Testing
└── results
└── iSeg2017
└── iSeg-2017-Testing
The code has been tested with the following configuration
- h5py == 2.7.0
- ipython == 5.3.0
- jupyter == 1.0.0
- keras == 2.0.2
- nibabel == 2.1.0
- nipype == 0.12.1
- python == 2.7.12
- scipy == 0.19.0
- sckit-image == 0.13.0
- sckit-learn == 0.18.1
- tensorflow == 1.0.1
- tensorflow-gpu == 1.0.1
Once all the libraries above have been installed, the following step is to run the jupyter notebook on the folder containing the iSeg2017.ipynb file.
jupyter notebook