Implementation of VoxResNet

This is a repository containing code of Deep Voxelwise Residual Networks for Volumetric Brain Segmentation (VoxResNet) [1].

Note that this is not an official implementation.

Requirements

• python 3.6
• chainer v2
• cupy
• dipy
• nibabel
• numpy
• pandas
• SimpleITK

Preparing dataset

1. Download Internet Brain Segmentation Repository (IBSR) (or other dataset you want to try)
2. preprocess dataset
   • training dataset $ python preprocess.py -i /path/to/IBSR/dataset -s IBSR_01 IBSR_02 IBSR_03 IBSR_04 IBSR_05 --input_image_suffix _ana_strip.nii.gz --output_image_suffix _preprocessed.nii.gz --label_suffix _segTRI_ana.nii.gz -f dataset_train.json --n_classes 4 --zooms 1. 1. 1.
   • test dataset $ python preprocess.py -i /path/to/IBSR/dataset -s IBSR_11 IBSR_12 IBSR_13 IBSR_14 IBSR_15 --input_image_suffix _ana_strip.nii.gz --output_image_suffix _preprocessed.nii.gz --label_suffix _segTRI_ana.nii.gz -f dataset_test.json --n_classes 4 --zooms 1. 1. 1.

Train VoxResNet

$ python train.py -g 0 -f dataset_train.json

Test VoxResNet

$ python segment.py -g 0 -i dataset_test.json -m vrn.npz -o _segTRI_predict.nii.gz

Reference

[1] Chen, Hao, et al. "VoxResNet: Deep Voxelwise Residual Networks for Volumetric Brain Segmentation." arXiv preprint arXiv:1608.05895 (2016).