Implementation of a data augmentation (DA) approach with the aim of reducing the scanner bias of models trained on single-scanner data. If you use this code please cite our paper:
Meyer, M.I., de la Rosa, E., Pedrosa de Barros, N., Paolella, R., Van Leemput, K., Sima, D.M. A Contrast Augmentation Approach to Improve Multi-Scanner Generalization in MRI. Front. Neurosci. 15:708196. doi: 10.3389/fnins.2021.708196
The method aims to increase the intensity and contrast variability of a single-scanner dataset such that it is representative of the variability found in a large multi-scanner cohort. To do this we randomly modify the MRI tissue intensities using a Gaussian Mixture Model based approach. As a result the contrast between tissues varies, as seen when different scanners are used at acquisition.
The following packages are necessary to run the code:
- python>=3.7
- numpy>=1.18
- matplotlib=3.3
- scikit-learn=0.23
- nibabel=3.2
The method is exemplified in the demo.py script. To run this you will need a T1 image and the corresponding brain
mask in nifti format (nii.gz).
We do not provide images to test the model on. The example results we show were
acquired on the publicly available OASIS dataset [1].
You can run this using all the default options by simply doing
python demo.py -i <path to T1> -m <path to mask>
Optional arguments include:
-n number of components in the mixture (default: 3)
-mu range of variability to change the mean of each component in the mixture (default: [0.03 0.06 0.08])
-s range of variability to change the standard deviation of each component in the mixture
(default: [0.012 0.011 0.015])
-p percentiles to use when clipping the intensities. (default [1 99]).
The default range of variability to change the values of the components were estimated on a real clinical dataset, and are used as default to represent the method as submitted to ISBI. For the demo, if you wish to use a different number of components you also need to define the new values of -mu and -s. Let's say you want to run the method using 4 components, to allow the mean and standard deviation of each component to change by (0.1, 0.2, 0.3, 0.4) and (0.01, 0.02, 0.03, 0.04), and to clip the intensity percentiles at (0.1, 99.9):
python demo.py -i <path to T1> -m <path to mask> -n 4 -mu 0.1 0.2 0.3 0.4 -s 0.01 0.02 0.03 0.4 -p 0.1 99.9
This script saves a nifti image to a results/ folder, which will be created automatically if it does not exist.
A comparison to the original image is also plotted and saved for inspection.
The method performs better if images have been bias-field corrected, but this is not strictly necessary. When exploring the method it is also possible to set a fixed variation term for each of the components (see function generate_gmm_image() in augmentation_utils.py).
[1] Marcus, DS, Wang, TH, Parker, J, Csernansky, JG, Morris, JC, Buckner, RL. Open Access Series of Imaging Studies (OASIS): Cross-Sectional MRI Data in Young, Middle Aged, Nondemented, and Demented Older Adults Journal of Cognitive Neuroscience, 19, 1498-1507. doi: 10.1162/jocn.2007.19.9.1498