Region of Interest

Regions of Interest in fMRI

Regions of interest (ROIs) are specific areas of the brain that are selected as a statistical map of that region. There are several reasons for creating ROIs. They are useful to study a particular brain area, to identify patterns of activity for that area and to control for type I error by limiting statistical tests (see the problem of multiple testing).

ROI for exploration

If your are interested in functionally defining new ROIs, the best way is to extract the activity pattern induced by some condition. You can then use a searchlight algorithm or a thresholded activation map to extract the peaks of activation from a predefined sphere. In this way it is possible to locate peaks of activity whose coordinates can be used for further studies. However, it is not possible to perform statistical tests within such ROIs as they have been defined by significant values from a specific activity map.

ROI for statistical control

Statistical inference is only possible if the locations of the ROIs are predefined. For example, several anatomical ROIs and can be selected and compared via statistical tests. This reduces the total number of statistical tests required, but it is not clear whether such an approach is better than Gaussian random field theory or false discovery rate corrections. Therefore, SPM uses voxel-wise analysis for a set of ROIs and controls for multiple comparisons in these voxels using Gaussian random field theory.

ROI for functional specification

The same approach can be used for functional ROIs. However, these need to be defined by a different contrast than e.g. localisation scans. In this way, ROIs can be identified from a chosen contrast and their location can be used to detect activity within a functional ROI. Thus the regions are no longer bound by some anatomical specificity but by functionality alone.

Defining ROIs

When defining ROIs it is adviced to to so for each individual subject, based on their anatomy. This is important because there is a large intersubject variability in ROI locations. However one can use single-subject anatomical atalses, such as the AAL atlas or the Talairach atlas. Best is to apply probabilistic atlases of macrosopic anatomy or of Brodmann's areas such as implemented in the SPM Anatomy Toolbox. Another way of narrowing down ROI locations might be reliable brain coordinates found in previous studies, preferably meta-analyses.

Resource

Poldrack R. A. (2007). Region of interest analysis for fMRI. Social cognitive and affective neuroscience, 2(1), 67–70. https://doi.org/10.1093/scan/nsm006