Using Brats 2015 dataset
As follows make diretories. Unzip brats dataset in data directory.
U-net3D
data
HGG LGG HGG_LGG output source
- Run data_rename.py in U-net3D directory.
This will rename brats dataset.
- Run mha_to_nii.py
This will change MRI ground truth file extension(.mha to .nii) and move file to /data/data directory
- Using 3D slicer for doing N4ITKBiasFieldCorrection.
N4N4ITKBiasFieldCorrection remove image gradation in your MRI dataset.
cd /U-net3d/data
for n in *.mha; do ~/slicer/lib/Slicer-4.7/cli-modules/N4ITKBiasFieldCorrection "./$n" ./data/"${n%.mha}.nii"; done
3D slicer's file root maybe diffenent.
- Run pre_process.py
This source contains three pre-processes
Intensity Range Standardization Histogram equalizing Gauss Normalization
Run train.py
Run predict.py
- 윤지석, 석홍일 (2016). 딥러닝 기반의 멀티-모달 MRI 영상에서의 뇌종양 영역 분할. 한국정보과학회 학술발표논문집, 1680-1682
- Kayalibay, B., Jensen, G., & van der Smagt, P. (2017). CNN-based segmentation of medical imaging data. arXiv preprint arXiv:1701.03056.