Human Brain Mapping / Comparison of continuously acquired resting state and extracted analogues from active tasks

Functional connectivity analysis of brain networks has become an important tool for investigation of human brain function. Although functional connectivity computations are usually based on restingstate data, the application to taskspecific fMRI has received growing attention. Three major methods for extraction of restingstate data from taskrelated signal have been proposed (1) usage of unmanipulated task data for functional connectivity; (2) regression against task effects, subsequently using the residuals; and (3) concatenation of baseline blocks located inbetween task blocks. Despite widespread application in current research, consensus on which method best resembles restingstate seems to be missing. We, therefore, evaluated these techniques in a sample of 26 healthy controls measured at 7 Tesla. In addition to continuous restingstate, two different task paradigms were assessed (emotion discrimination and right fingertapping) and five welldescribed networks were analyzed (default mode, thalamus, cuneus, sensorimotor, and auditory). Investigating the similarity to continuous restingstate (Dice, Intraclass correlation coefficient (ICC), R2) showed that regression against task effects yields functional connectivity networks most alike to restingstate. However, all methods exhibited significant differences when compared to continuous restingstate and similarity metrics were lower than testretest of two restingstate scans. Omitting global signal regression did not change these findings. Visually, the networks are highly similar, but through further investigation marked differences can be found. Therefore, our data does not support referring to restingstate when extracting signals from task designs, although functional connectivity computed from taskspecific data may indeed yield interesting information.(VLID)484322