General purpose medical segmentation

This program can segment liver tissue for 2D CT slices and 3D MR volumes using the deep learning technique. This method based on UNet++ architecture integrates the probabilistic map into the encoder of network. The network is trained in an iterative fashion, since the probabilistic map comes from the previous classifier.

data augmentation

augmentation.py dataset.py

model

Net.py loss.py

running

train.py

data

The data folder includes five MR volumes(imgx.nii.gz) and their corresponding ground truth(labelx.nii.gz).