- localization and quantification of artifacts (motion, magnetic field inhomogeneity and noise) in binary or multi-class setting
- correction of motion-induced artifacts (rigid and non-rigid motion)
- visualize the trained kernels and feature maps
- deep visualization: significance map of trained network content, backpropagate most-likely input patch and sparse attractor points of a test image
easy-to-use graphical interface for medical deep learning
- 2D/3D data viewer
- data preprocessing: labeling, patching, data augmentation, data splitting
- network training: parameter setting, training/validation/test set selection, call to DL backend (keras, Tensorflow, ...)
- test data evaluation: accuracy/loss plots, confusion matrix and derived metrics
- network visualization: kernel weights, feature maps and deep visualization
Install the requirements
$ python3 -m pip install -r requirements.txt- define database layout in
config/database/_NAME_OF_DATABASE_.csv(as specified in param.yml -> MRdatabase) - edit parameters in
config/param.yml - run code via
main.py
training/prediction can also be invoked from the GUI. Please adapt mainGUI_Template.py according to your needs
Qt_main.py
main.py ==> model.fTrain()/fPredict()
| Network | Artifact type detection | Publication |
|---|---|---|
| CNN2D | motion_rigid motion_non-rigid motion_both |
1, 7 |
| CNN3D | motion_rigid motion_non-rigid motion_both |
2, 6 |
| MNetArt | motion_rigid motion_non-rigid motion_both |
2, 4 |
| VNetArt | motion_rigid motion_non-rigid motion_both |
2, 4, 5 |
| DenseNet | motion_both inhomogeneity noise |
|
| DenseResNet | motion_both inhomogeneity noise |
3 |
| ResNet | motion_both inhomogeneity noise |
|
| GoogleNet | motion_both inhomogeneity |
|
| InceptionNet | motion_both inhomogeneity noise |
3 |
| VGGNet | motion_both inhomogeneity |
- Küstner, T., Liebgott, A., Mauch, L., Martirosian, P., Bamberg, F., Nikolaou, K., Yang B., Schick F. & Gatidis, S. (2017). Automated reference-free detection of motion artifacts in magnetic resonance images. Magnetic Resonance Materials in Physics, Biology and Medicine, 1-14.
- Küstner, T., Jandt, M., Liebgott, A., Mauch, L., Martirosian, P., Bamberg, F., Nikolaou, K., Gatidis, S., Schick, F. & Yang, B. (2018). Automatic Motion Artifact Detection for Whole-Body Magnetic Resonance Imaging. Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).
- Küstner, T., Liu, K., Liebgott, A., Mauch, L., Martirosian, P., Bamberg, F., Nikolaou, K., Yang, B., Schick, F. & Gatidis, S. (2018). Simultaneous detection and identification of MR artifact types in whole-body imaging. Proceedings of the International Society for Magnetic Resonance in Medicine (ISMRM).
- Küstner, T., Jandt, M., Liebgott, A., Mauch, L., Martirosian, P., Bamberg, F., Nikolaou, K., Gatidis, S., Yang, B. & Schick, F. (2018). Motion artifact quantification and localization for whole-body MRI. Proceedings of the International Society for Magnetic Resonance in Medicine (ISMRM).
- Liebgott, A., Milde, S., Jandt, M., Mauch, L., Martirosian, P., Bamberg, F., Schick, F., Nikolaou, K., Yang, B., Gatidis, S. & Küstner, T. (2018). Impact of Labeling Process on Automated Motion Artifact Detection in Whole-Body MR Images with a Deep Learning Approach: A Comparative Study. Proceedings of the ISMRM Workshop on Machine Learning.
- Küstner, T., Liegbott, A., Mauch, L., Martirosian, P., Schick, F., Bamberg, F., Nikolaou, K., Yang, B. & Gatidisi, S. (2017). Automatic reference-free motion artifact detection and quantification in T1-weighted MR images of the head and abdomen. Proceedings of the Annual Scientific Meeting (ESMRMB).
- Küstner, T., Liebgott, A., Mauch, L., Martirosian, P., Nikolaou, K., Schick, F., Yang, B. & Gatidis, S. (2017). Automatic reference-free detection and quantification of MR image artifacts in human examinations due to motion. Proceedings of the International Society for Magnetic Resonance in Medicine (ISMRM).