This is the open source toolkit for the MBIA-MICCAI 2019 paper titled Structural Similarity based Anatomical and Functional Brain Imaging Fusion by N.Kumar et al.
The paper performs the following tasks:
- Develops a novel end-to-end unsupervised fusion algorithm for anatomical and functional imaging modalities using Convolutional Neural Networks.
- Develops a mathematical framework to visualise the fused image using the partial derivatives with respect to the input imaging modalities.
- Compares the proposed fusion + visualisation method with other state-of-the-art fusion methods.
Note: Please cite the paper if you are using this code in your research.
- Python 2.7
- Tensorflow 1.0.x or Pytorch 1.0 and above
- numpy
- matplotlib
- skimage
- MATLAB
The training of the FunFuseAn network was done with 500 MRI-PET image pairs available at Alzheimer’s Disease Neuroimaging Initiative (ADNI) as mentioned in the paper. Although, the data is available for public use, it is still required to apply for getting the access to the repository by filling out a questionaire. In case you are interested to obtain the data, please apply at this link. For conducting the inference of our network, in addition to the ADNI test data, we also used 10 image pairs from Harvard Whole Brain Atlas which is also publicly available and donot required any written permission. We therefore uploaded this data for your usage.
For training and testing the FunFuseAn network, you can use either Tensorflow or PyTorch as your Deep learning package while we only provide Tensorflow code for the visualisation part. For easier implementation, we recommend you to follow the jupyter notebooks (for both tensorflow and pytorch versions) provided in this repository.
As mentioned in the paper, we quantitatively evaluated the quality of our end-to-end unsupervised learning based medical image fusion method with some popular fusion metrics using MATLAB. The code of the fusion metrics used has been provided in the folder 'Fusion metrics' in this repository. The author and citation details of each of these metrics are given in the comment section of the code. If you use these fusion metrics for evaluation of your own fusion method, we recommend you to properly cite the original contribution. For running these metrics, you need two grayscale input images and a grayscale fused image. For getting the metric scores all at once, you need to run fusionAssess.m file.
As mentioned in the paper, we used several existing fusion methods and compared their metric scores with those obtained from FunFuseAn. The code for the evaluated fusion methods are publicly available. In case of any questions related to these fusion methods, please contact the corresponding authors.
@InProceedings{10.1007/978-3-030-33226-6_14,
author="Kumar, Nishant
and Hoffmann, Nico
and Oelschl{\"a}gel, Martin
and Koch, Edmund
and Kirsch, Matthias
and Gumhold, Stefan",
title="Structural Similarity Based Anatomical and Functional Brain Imaging Fusion",
booktitle="Multimodal Brain Image Analysis and Mathematical Foundations of Computational Anatomy",
year="2019",
publisher="Springer International Publishing",
address="Cham",
pages="121--129",
isbn="978-3-030-33226-6"
}