Inter-modality co-registration

There are two possible ways to initiate inter-modality co-registration:

• Choose to co-registration option from within a specific analysis module (e.g. BOLD fMRI, DTI, ASL etc).
In nordicICE workspace: Drag one data set onto another while pressing down the c -key on your keyboard.

Either way, the MPR view will open with the source volume as overlay onto the reference volume, and with the co-registration panel on the right-hand side. Note that the first volume in the series will be used as source/reference volume if the data is 4-D.

The user has the option to choose either manual or automatic co-registration, or even combine them by manually selecting the starting point for the automatic co-registration. The two data sets that are to be co-registered are shown as thumbnails at the top of the coreg panel, with the source data to the left and the reference data to the right. The correctness of the co-registration is verified by visual inspection, and the co-registration information is then saved to the database if the user chooses to do so.

Automatic co-registration

Automatic inter-modality co-registration uses a mutual information-based algorithm to search an optimal rigid transformation that aligns the two data sets. The implementation is based on an article by H. Sundar et. al. titled "Robust computation of mutual information using spatially adaptive meshes" . The main point in this article is to only consider a subset of the voxels in the computation, i.e. voxels that represent the non-homogenous areas.

Both volumes are resampled onto the same uniform grid. This grid has the same orientation and spans the same area as the reference volume, unless functional to structural co-registration is applied. In that case, the roles are changed because functional data normally covers a smaller part of the brain compared to the structural data, and we would not want to include non-overlapping areas.

Some parameters can be edited by the user, and these are:

Voxel size: This parameter determines the distance between grid points along each of the axes.
Smooth volumes / width: The match between the two volumes is expressed as a cost function, and the problem of finding the optimal transformation is equivalent to finding the minimal value of this cost function. Smoothing can help avoiding locally minima, but could also cause misalignment because information in the pixel values is lost. Gaussian smoothing is used, and the filter width is given by the width parameter (in voxel units).
Expand grid / size: If there is a large mismatch between the reference and the source volume, it would help to expand the uniform grid. The size of the grid is expanded by the given factor in each direction.
Sub-sample / levels: Sub-sampling is like smoothing, but with the benefit that computation time is much lower on the sub-levels because of the reduced number of grid points.
Histogram: The number of histogram bins. The histogram is used in the calculation of the mutual information cost function.
Smooth histogram: Has the same effect as choosing a smaller number og histogram bins.
Brain extraction: When co-registering functional to structural data, the results are much more accurate when the scull is stripped from the structural volume.

Manual co-registration

The user interface for manual coregistration uses much of the functionality previously described to the ordinary MPR. One thing to note immediately, however, is that all transformations (translations, rotations and scales) applied in manual coregistration are applied to the overlay (source) volume only. The underlay image volume is untouched.

When manual coregistration is launched additional orientation labels occur in the projection panels. The added labels are smaller and written in italics to separate them from the ones already present in MPR. These italicized labels refer to the orientation of the overlay. Thus, as the overlay volume is transformed, the overlay orientation labels are updated while the underlay labels remain fixed. The same rule for shading as for the underlay orientation labels are applied to the overlay labels.

The coregistration panel has a matrix for numerical input of transformation parameters. The columns are used to input the amount of translation, rotation, scale and the position of the (rotation and scale) vertex. The the first three rows allow separate parameter input for each direction, i.e. right-to-left (R->L ), anterior-to-posterior (A->P ) and inferior-to-superior (I->S ). The fourth row control the amount of increment and decrement observed when clicking the parameters up and down buttons in the corresponding collumn. Below the increment/decrement row, there are reset buttons for each type of transformation.

Visualization panel

The lower part of the coregistration panel contains a group of visualization controls. These controls customize the opacity and the palette of the overlay, whether the overlay is presented as a checker board, the checker board resolution, the hidden state of the overlay and finally whether transformations are previewed. Many of these options are also accessible in the Overlay settings dialog window as described in the section on Modify image overlays .