Dear Dr. Jung,

Thank you for your positive feedback about the toolbox. Regarding the removal of confounding effects in the conn toolbox, if you keep the default settings the toolbox will perform (in this order) despiking, nuisance regression/removal (including motion and CompCor components, as well as a linear detrending term), and band-pass filtering. You may remove any of these steps by changing the selected options in the conn toolbox 'Preprocessing' tab, but you can not change their ordering. Detrending (when selected) is always implemented as part of nuisance regression/removal step by adding a linear term as an additional regressor to the general linear model. If in addition you select the 'Simult' option, then all of the regressors in this general linear model will be filtered (using the same band-pass filter as that apply to the data) before performing the regression step, which effectively implements a simultaneous filtering-regression operation (it is a different implementation but it leads to exactly the same results as if one were to add a set of low-frequency and high-frequency regressors into the general linear model, and then performing a single regression step to simultaneously remove the nuisance factors and perform band-pass filtering). I added a small discussion about the difference between the 'Simult' and the default 'RegBP' procedures in this news post if you are interested (http://www.nitrc.org/forum/forum.php?for...), and you can read about these issues in more detail in this recent manuscript (Hallquist, M.N., Hwang, K., Luna, B., 2013. The nuisance of nuisance regression: Spectral misspecification in a common approach to resting-state fMRI preprocessing reintroduces noise and obscures functional connectivity. NeuroImage, 82:208-225)

Regarding the potential application of despiking before the spatial preprocessing steps, I do not know of any studies looking at the potential differences here, but my impression is that as long as you do not perform slice timing correction (and the standard recommendation for rsfMRI is that this step is typically not necessary since the frequencies of interest in rsfMRI are often considerably lower than the timing differences introduced by the slice acquisition order) the two procedures (despiking before or after spatial preprocessing) should lead to very similar results for most standard sources of BOLD signal artifacts (those that typically benefit from despiking as a way to minimize their effect on the functional connectivity measures). Another related procedure that you may be interested in is 'scrubbing', which attempts to identify potential outlier scans and remove them from the analyses before computing functional connectivity measures. In the conn toolbox this is not yet part of the default behavior but you can still easily implement this using ART (http://www.nitrc.org/projects/artifact_d...) as an additional preprocessing step and then including the ART-generated SPM-regressors as additional first-level covariates (and confounding effects) into the conn toolbox in order to remove the identified scans from the analyses. 

Hope this helps
Alfonso




Originally posted by Wi Hoon Jung: