Basic Information of Code
This is an implementation of a simple neural network to segment brain tissues (white matter, gray matter and cerebrospinal fluid).
Data of brain volume is downloaded from BrainWeb.
Author: Qixun Qu
Date: 2017/05/02
Dependencies
- python 3.5.2
- numpy 1.11.3
- nibabel 2.1.0
- tensorflow 1.0.0
- tensorlayer 1.4.1
- matplotlib 2.0.0
All functions are tested in Ubuntu 16.04.
Run the Segmentation
Three steps to run the program.
1. Download brain volume from BrainWeb, including ground truth, T1, T2 and PD data in .mnc files. Put these files into folder "Data".
2. In terminal, run "python seg_brain.py" to train the model. The model will be save in file model.npz.
3. In terminal, run "python test.py" to test a slice, segmentation results will be displayed, Dice index and Jaccard index is computed as well.
Results
Loss and dice index.
Segmentation result of one slice.
