Improving the balance between sensitivity and specificity for functional localization in fMRI data

In fMRI, localizer tasks are often used to define and examine the activity of functional regions of interest (fROIs). Current statistical methods tend to localize fROIs inconsistently, only guarding against false positives while avoiding missing true activation is equally important in this context. In recent work, we studied alternative based thresholding (ABT, Durnez et al., 2013) that allows taking into account an effect size (ES) that reflects practically relevant activation. The combination of evidence against the null of no effect and evidence against the alternative ES results in a layered statistical map that displays evidence for activation in brain regions. To deal with large-scale testing in fMRI, Kang et al. (2015) proposed a likelihood ratio approach that contrasts evidence of true activation against evidence of the null. We study this approach in the context of fROIs and show how a combination of ABT and the likelihood approach can lead to optimal decision criteria for defining fROIs. The achieved balance between sensitivity and specificity can be user-specified providing decision tools that meet the different needs of researchers in the field.

http://ift.tt/2pBBSFC