Module: Register Images ()

Description:

This module computes an affine transformation for the co-registration of two image data sets, using an iterative optimization algorithm. A hierarchical strategy is applied, starting at a coarse resampling of the data set, and proceeding to finer resolutions later on. Different similarity measures like Euclidean distance, mutual information, and correlation can be chosen.

To use this module, you must connect it to two scalar fields. To follow the progress visually, a module like Bounding Box or Isosurface should be connected to each of them, but this may slow down the registration process. If the Apply button is pressed, the module starts by successively optimizing the transformation of the first input.

The optimization can be interrupted at any time. Interruption might take some seconds.

For more information, see the tutorials about registration of 3D data sets, in particular More about the Register Images module section.

Note: The Normalized Mutual Information metric is based on the following publication:
Studholme, C., Hill, D.L.G., Hawkes, D.J.: An Overlap Invariant Entropy Measure of 3D Medical Image Alignment. In: Pattern Recognition 32 (1999), S. 71-86

Further references include:
Viola, P.A.: Alignment by Maximization of Mutual Information, Massachusetts Institute of Technology, Diss., 1995

Collignon, A., Maes, F., Delaere, D., Vandermeulen, D., Suetens, P., Marchal, G.: Automated Multi-modality Image Registration Based on Information Theory. In: IPMI 1995. Dordrecht, Niederlande: Kluwer Academics, 1995, S. 263-274

and

Pluim, J.P.W., Maintz, J.B.A., Viergever, M.A.: Mutual-Information-Based Registration of Medical Images: A Survey. In: IEEE Trans. MI 22 (2003), S. 986-1004

Note: The Correlation Ratio metric is explained in the following publication:
Roche, A., Malandain, G., Pennec, X., and Ayache, N.: The Correlation Ratio as a New Similarity Measure for Multimodal Image Registration. INRIA Sophia Antipolis, EPIDAURE project, 1998

Connections:

Model [required]
The model data set to be transformed.

Reference [required]
The reference data set to which the model is registered.

Reference2 [optional]
See Reference3 below.

Reference3 [optional]
The connection ports Reference2 and Reference3 are used for a particular type of images routinely taken in MRI, so-called Localizers. A Localizer consists of few (~3) slices each of an axial/xy, sagittal/yz, and coronal/yz scan. A CT or MR taken from the same patient or object can be registered to the Localizer images by using high resolution information for each direction in separate volumes. To do so, connect the two additional Reference ports with the Localizer images. Note that the port Options2: Localizer automatically gets checked.

Ports:

Transform

At this port you can select the number of transformation parameters to be optimized. The numbers are 6 for Rigid (3 translations and 3 rotations), 7, 9, and 12 for IsoScale, AnisoScale, and Shear, respectively.

If you select for example Rigid and AnisoScale, on each resolution level the 6 parameters of a rigid transformation will be optimized first, followed by an optimization of all 9 parameters of an anisoscale transformation. In this way as much as possible is done using a rigid transformation.

If only AnisoScale is selected, all 9 parameters will be optimized at once. A larger contribution of the scaling parameters may be expected for this selection.

Disable Rotation

If this option is checked, the transform used will have no rotation part. If the Shear option of the Transform port is enabled this feature will be automatically disabled.

Register

Select here the registration mode. If 2D is checked, the search is restricted to transformations within the same plane. In the case 3D is checked a full 3D registration is performed.

Threshold outside [advanced]

Depending on your application, the final model position can have different overlaps with the reference data set. For example, the model can lie completely inside the reference, or it can share only a small overlapping area. Overlapping area refers to the space that is common to the bounding boxes of both data sets. This port lets you use your prior knowledge about the overlap. Its values are between 0 and 1. If you expect that the model and the reference should have a large overlap, then set a large value. Due to numerical errors, it is not recommended to set the threshold value too close to 1. Values around 0.8 should be suitable. Otherwise, if you know that probably only small parts overlap, then use a small value or even zero. The higher the value, the stronger is the decrease of the metric value if the model has parts outside the bounding box of the reference.

Prealign

If the Align Centers button is pressed, the centers of gravity of both data sets are computed, taking the image intensity as a mass density. The model data set is translated in order to align both centers of gravity.

Metric

Metric

This port selects the similarity measure to be applied.

• Euclidean means the Euclidean distance, i.e., the mean squared difference between the gray values of model and reference. The more the images are close to each other, the more the metric value will approach 0.
• Correlation measures the correlation ratio of the registered images. It takes values between 0 and 1. The case correlation ratio = 1 correspond to the case of two identical and registered images.
• The Mutual information metrics, especially the normalized one, are recommended when images from different modalities, e.g., CT and MRT, are to be registered.
• The Label Difference measure difference between two connected labels.

The metric will affect the result of the registration, since the metric is the factor which determines if the images are aligned or not.

For example, images which come from different modalities will have a bad similarity value if the euclidean metric is used, whatever the position of the model and the reference. Using Euclidean for these images won't be a good choice. The Mutual Information can be used instead.

For more information, see More about Register Images > Similarity metrics

Histogram range reference [advanced]

This port is only active if one of the Mutual information metrics, or the Correlation metric has been selected. Here you can define the range of gray values for the reference data set. For the Mutual information metrics, gray values outside the range are sorted into the first or last histogram bin, respectively. For the Correlation metric, voxels with gray values outside the range are ignored. It is a good idea to determine the range via a visualization of the data set, e.g., using an Ortho Slice module.

Histogram range model [advanced]

The same as the previous port, now for the model data set.

Histogram bins [advanced]

For the Normalized Mutual Information andMutual Information metric it is possible to set the number of bins for model and reference histograms The number of bins basically determines the resolution by which the range of gray values set in ports Histogram range reference/model is partitioned. The more bins the more accurate the gray values are resolved but the more computation time is needed. The default values set by the module are usually a good compromise.

Compute [advanced]

If the Metric value button is pressed, the current metric value is computed (depending on the current model position) and output on the console.

Resampling Options

CoarsestResampling [advanced]

At this port you can define the resampling rate for the coarsest resolution level where registration starts. The resampling rate refers to the reference data set. If the voxels of the reference data set are anisotropic, i.e., have a different size in x-, y-, and z-direction, the default resampling rates are chosen in order to achieve isotropic voxels on the coarsest level. If the voxel sizes of model and reference differ, the resampling rates for the model are chosen in order to achieve similar voxel sizes as for the reference on the same level.

Options [advanced]

If toggle IgnoreFinestLevel is selected, registration will be performed on all but the finest (i.e., the original) resolution. In many cases a sufficient accuracy can be achieved in this way. Registration on the finest level may slightly improve the accuracy, but the computation time will typically increase by an order of magnitude.

Optimizer Options

Optimizer Type [advanced]

At this port you can choose between different optimization strategies. The ExtensiveDirection or BestNeighbor optimizers are well suited for coarse resolution levels, the QuasiNewton or LineSearch optimizers for the finer resolution levels. The default strategy uses the ExtensiveDirection optimizer on the coarse levels and the QuasiNewton optimizer on the finest levels.

Optimizer step [advanced]

This port sets the initial and the final value for the stepwidth to be applied in the optimizations. These stepwidths refer to translations. For rotations, scalings, and shearings appropriate values are chosen accordingly.

The default value for the initial stepwidth is 1/5 of the size of the bounding box. If both data sets are already reasonably aligned, you may choose a smaller initial stepwidth.

The default value for the final stepwidth is 1/6 of the voxel size.

Gradient optimizer [advanced]

At this port you can select the number of resolution levels (between 0 and 2) where the QuasiNewton optimizer is applied. On the coarser levels the optimizer as selected at port Optimizer is applied. If the number of levels is less than or equal to the number selected at this port, the optimizer as selected at port Optimizer is applied at least at the coarsest level.

Localizers

Localizers [advanced]

This port is automatically checked upon connecting ports Reference2 and/or Reference3 with localizer images. If checked Register Images will use those images during registration.

If the Align Principal Axes button is pressed, the centers of gravity and moments of inertia of both data sets are computed, again taking the image intensity as a mass density. The principal moments of inertia and corresponding principal axes are computed. The best of 24 possible alignments of the principal axes is determined according to the similarity measure as selected at the Metric port.

Pressing the Register button starts the actual registration process.