MoCoGUI
A Matlab toolbox which allows to perform, visualize and analyze different image registration algorithms. The toolbox is split into three GUIs:
- RegGUI: for performing image registration
- EvalGUI: for registration visualization and quantitative evaluation
- LandmarkGUI: for feature-based evaluation via landmark points, lines and ROIs
Please refer to the Documentation for more details.
RegGUI
Images (2D, 3D, 4D) can be loaded into the GUI from the Matlab workspace or from the hard drive. The supported file formats are:
- DICOM
- NIfTI
- MAT
- MHD/RAW
- GIPL
- TIFF, JPEG, BMP, PNG
The loaded images can then be coregistered onto each other via different registration algorithms:
- elastix [Klein et al., TMI 2010; Shamonin et al., Front Neuroinform 2010]:
- Parametric (rigid, affine, non-rigid) registrations with various solvers - http://elastix.isi.uu.nl/download.php
- HALAR/LREG (Hierarchical adaptive local affine registrations) [Buerger et al., MEDIA 2011]:
- hierarchical image splitting and parametric B-Spline registration of variable-sized blocks - https://www.isd.kcl.ac.uk/internal/hyperimage/
- LAP (Local All-Pass) [Thevenaz et al., TMI 2000; Blu et al., TMI 2001; Blu et al., TIP 2004; Gilliam et al., ISBI 2016]:
- Optical flow-based registration which reflects non-rigid deformations as local rigid displacements resp. local phase-shifts, i.e. local all-pass filter operations - https://sites.google.com/site/cwsgilliam/3D-LAP
- Demons [Thirion, MEDIA 1998]:
- Diffusion-based registration motivated by optical flow - http://www.mathworks.com/matlabcentral/fileexchange/21451-multimodality-non-rigid-demon-algorithm-image-registration
The GUI is easily extendable for further registration algorithms.
Transformed images and deformation fields are saved back to the hard drive.
EvalGUI
The obtained registration results can be visualized and evaluated in the EvalGUI. Retrieved quantitative evaluation results can be exported to the hard drive or Excel/CSV file for further analysis.
Visualization
Views from different orientations of the images and deformation fields, zooming and scrolling (via mouse wheel) through different slices is supported, as well as contrast/brightness adjustment.
Evaluation
The derived transformed images and deformation fields can be examined via:
- color-coded difference image
- determinant of Jacobian
- divergence of transformation
- deformation field magnitude
- gradient of deformation field
- overlap measures (Dice, Jaccard) of automatic segmented lung
- intensity-based similarity metrics (MI, SSD, SSIM, RMSE, ...)
LandmarkGUI
Anatomic landmarks (points, lines, ROIs) can be set to perform a feature-based evaluation of the image registration. Landmarks are set in the reference and transformed images to allow the calculation of overlap measures and distance metrics. The obtained overlap results are visualized in the GUI and exported to the hard drive for further analysis.
A quick guide describes the labeling procedure: LandmarkGUI Quick Guide
Download
The GUIs are publicly available: https://github.com/thomaskuestner/MoCoGUI
Please cite the paper, if you use it in a scientific publication.