Skip to content
master
Go to file
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
 
 
 
 
 
 

README.md

Multimodal Brain Synthesis

This project performs multimodal MR brain synthesis using modality invariant latent representations. For details see our papers Multimodal MR Synthesis via Modality-Invariant Latent Representation and Robust Multi-Modal MR Image Synthesis.

The main files in this project are:

  • model.py: contains the neural network implementation
  • loader_multimodal.py: loads the input data into a Data object and performs pre-processing.
  • runner.py: creates an Experiment object to perform cross validation on a given Data object.

An example, assuming usage of a Data object is the following:

data = Data(dataset='ISLES', trim_and_downsample=False)
data.load()

input_modalities = ['T1', 'T2', 'DWI']
output_weights = {'VFlair': 1.0, 'concat': 1.0}
exp = Experiment(input_modalities, output_weights, '/path/to/foldername', data, latent_dim=16, spatial_transformer=True)
exp.run(data)

The code is written in Keras and expects image_data_format to be set to channels_first.

Data

The sources used to evaluate our methods are brain data from ISLES and BRATS challenges, as well as the IXI dataset.

Citation

If you use this code for your research, please cite our papers:

@article{chartsias2018multimodal,
  title={Multimodal MR Synthesis via Modality-Invariant Latent Representation},
  author={Chartsias, Agisilaos and Joyce, Thomas and Giuffrida, Mario Valerio and Tsaftaris, Sotirios A},
  journal={IEEE Transactions on Medical Imaging},
  year={2018},
  volume={37},
  number={3},
  pages={803-814},
  publisher={IEEE},
  doi={10.1109/TMI.2017.2764326},
  ISSN={0278-0062}
}

@inproceedings{joyce2017robust,
  title={Robust Multi-modal MR Image Synthesis},
  author={Joyce, Thomas and Chartsias, Agisilaos and Tsaftaris, Sotirios A},
  booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
  pages={347--355},
  year={2017},
  organization={Springer}
}

Acknowledgements

The project uses a Spatial Transformer implementation, distributed under MIT licence.

About

No description, website, or topics provided.

Resources

License

Releases

No releases published

Packages

No packages published

Languages

You can’t perform that action at this time.