Time Series Image Registration

Use the Time Series Registration tool when you have images which are already fairly closely aligned, such as occurs with fMRI, perfusion, or DCE-MRI images. The Time Series Registration tool is optimised to perform registration more quickly when only a small transform is needed; this is possible because the search space for optimal transforms is restricted. The other difference between the Time Series Registration tool and normal registration is that there is no explicit "base image" to which images are registered. The base image is implicitly defined by averaging all of the input images before registration. This average image is used as the "base image".

To start the Time Series Registration tool, select it from the Toolkits menu:

This brings up the Time Series Registration tool:

Selecting the Images to Register

The time-series of images to be registered can be organised in one of three ways:

Multiple input images (one per time point). You have as many input images are there are time points. Each image contains all slice locations that you want to analyse.
Multiple input images (one per slice location). You have as many input images are there are slice locations. Each image contains all time points.
Single input image (all time points and all slices). You have just one input image. This image contains all the data (all slices and time points) to be analysed. If this is the case, then you must also tell the tool about the ordering of data within this image:
- Time points are contiguous in image file. Select this option if the ordering of slices has all time points for the first slice location followed by all time points for the second slice location, etc.
- Different slice locations are contiguous in image file. Select this option if the ordering of slices has all slice locations for the first time point followed by all slice locations for the second first time point, etc.
Select the single input image by clicking on the and using the file chooser to select the input image, or by pressing the right mouse button and selecting from the menu of recently-used images.
You must also tell the tool how many time points are in the image so that it can distinguish time points from slice locations.

If you selected either of the first two options (multiple input images), then the window will look like below:

The panel below the "Input images:" label shows the current list of selected input images. You use the:

button to select a new set of images and put them to this list.
button to select a some more images and add them to this list.
button to remove all the images from the list.
button to remove the selected the image(s) from the list (you must first click image(s) in the list to select them).
and buttons to move the selected image(s) up or down the list. Images should be in the correct order for analysis - either in the correct slice order, or in the correct time order. Use these buttons to reorder them if they are not.

If you selected the third option (single input image), then the window will look like below.

You will then need set set:

Different time points are contiguous in image file. Select this option if the ordering of slices has all time points for the first slice location followed by all time points for the second slice location, etc.
Different slice locations are contiguous in image file. Select this option if the ordering of slices has all slice locations for the first time point followed by all slice locations for the second first time point, etc.

Select the single input image by clicking on the

and using the file chooser to select the input image, or by pressing the right mouse button and selecting from the menu of recently-used images.

You must also tell the tool how many time points are in the image so that it can distinguish time points from slice locations.

Setting the Degrees of Freedom

The setting of the "degrees of freedom" (DoF) determines the what types of change (distortion) may be made to the registered images to make them align. Set the degrees of freedom as you would for the regular Registration tool.
Note: you cannot set the initial transform parameters for the Time Series Registration tool; an Identify transform is always used as the initial transform.

Other Settings

Other settings are as for the Registration tool. However, note that there is an extra option for the cost function:

. When selected, the edge-matching panel becomes visible:

When you include edge-matching, you define a set of regions of interest over which matching takes place. The cost function is only evaluated over that part of the image which is enclosed by ROIs. In addition, the cost function is multiplied by an extra term which is calculated by finding the image intensity gradient at the edge of the feature which is defined by the ROIs. The scalar product of the intensity gradient and the feature surface normal vector is calculated and integrated over the whole of the surface. This is the "edge matching" term by which the cost function is multiplied.

To include the edge matching, you should:

Load one of your input images into Jim's main display. Choose an image where the feature you want to register is well-defined.
Start the ROI Toolkit.
If your input image has all slices and all time-points, choose the time point at which you want to outline the feature.
Outline the feature on every slice on which it appear, but at just one time-point.
Save the ROIs to disk.

The Time Series Registration tool will attempt to determine the time-point at which you defined the ROIs. However, it it can't, it will pop-up an error message and you will need to set the time point at which the ROIs were defined manually using:

Now set the ROI file you just created into the edge-matching panel above.

Note: registration using edge matching is experimental. We would welcome feedback about the usefulness of this functionality.

Pop-up Graphs

You can see plots of the registration transform parameters found by selecting the

check box. After time-series registration is complete, one or more graphs will pop up showing various components of the transforms that were found when registering. Shown below is a typical plot of the translation part of the affine registration transforms. If you include rotation and shear components in your degrees of freedom, you will see similar plots of these components.

You can write the values from these plots to a disk file, as for all plots, by selecting Write from the File menu:

Jim Home