top | previous

FreeSurfer Tutorial: Registration of functional and structural data (25 minutes of exercises)

In FreeSurfer, a parametric map of brain activations can be visualized as a color-coded overlay on a high resolution anatomical volume or surface. This presentation allows functional activations to be associated with the anatomical regions where they are observed. The functional dataset from which a brain activation map is derived and the anatomical volume may both be in the patient's or subject's native space, or anatomical and functional datasets may be warped into some standard morphological space. In either case, to generate a parametric map that will be registered to a corresponding anatomical volume the set of acquired functional volumes from which the map is derived must first be registered to that anatomical reference (and thus implicitly to surfaces derived from it). FreeSurfer includes different tools for manual and automatic registration of multiple datasets; each are described below.

1.0 Registration of multiple datasets

1.1 Manual and automatic registration

In the data pre-processing pipeline typical for fMRI analysis, registration of the functional volumes to a high-resolution anatomical reference is an important step.

Registering two datasets maps the coordinate frame in one dataset to that of the other, so that an anatomical feature in a functional acquisition maps to the same feature in the anatomical reference frame. In this example, we assume that the functional volumes have already been pre-processed through a number of steps including motion correction, so that each individual functional volume is in register with the others. Then, a few different options exist for determining the registration matrix that brings the functional volumes to spatial alignment the anatomical reference. Manual registration involves adjusting all the available degrees of freedom by hand while visually comparing the two volumes, until the functional volume appears in good register with the reference. Some methods for automatic registration are available as well. These exercises experiment with both registration approaches.

1.1.1 Manual registration of a functional volume to a structural volume using tkregister2

Two volumes may be manually registered by prescribing a transformation that brings one into alignment with the other. In many software packages, manual registration can be specified interactively (i.e. by mouse-clicking and dragging) or by manually specifying the elements of a transformation matrix. To experiment with manual registration, use the tool tkregister2. Tkregister2's GUI allows one volume to be translated, rotated, scaled and viewed relative to other datasets for this purpose. The following exercise will illustrate how datasets can be manually registered using FreeSurfer tools.

1.1.2 Automatic registration of a functional volume to a structural volume with spmregister

Since manual registration takes a lot of time and art to accomplish well, there are some useful tools for automatically registering a functional volume to a structural one. One such tool is spmregister, which registers a volume to its FreeSurfer anatomical volume using SPM's spm_coreg and creates a FreeSurfer file containing the registration matrix. The quality of the automatic registration can be inspected using tkregister2 to visualize the transformed volume relative to the FreeSurfer anatomical to which it was registered.

FsTutorial/FuncRegistration (last edited 2008-04-29 11:45:55 by localhost)