This repository contains the code employed in our work: "Unbiased Shape Compactness for segmentation", which has been accepted at MICCAI 2017 and awarded with the "Students travel award MICCAI".
A version of the paper has been submitted to ArXiv paper
The code is available only in matlab and python at the moment.
If you use this code for your research, please consider citing the original paper:
- Dolz J, Ben Ayed I, Desrosiers C. "Unbiased Shape Compactness for segmentation." 755--763, in Medical Image Computing and Computer Assisted Intervention − MICCAI 2017: 20th International Conference (2017)
To run it, in the matlab folder, just execute the following function:
runCompactness
Inside this function you can select which example from the paper you want to reproduce by assigning to the variable example one of the four options (AORTA_MRI, ESOPHAGUS_CT, RIGHTVENT_MRI, AORTA_CT). Compactness parameters are fixed as used in the paper.
The python implementation is a translation from the Matlab code, and wasn't used in the original paper. Some minor features are missing, but the results are the same. It requires:
- Python 3
- Numpy
- Scipy
- PyMaxflow
You can test it with ./runCompactness.py.
- Split admm main loop in several functions
- Add missing sanity tests after Laplacian update
- Update the visualisation to match the Matlab version