I have recently came across the XCP_d post processing pipeline, and intend to use it to denoise and postprocess my fMRIprep preprocessed resting-state and task fMRIs.
After reading through the documentation, I understand that for resting-state XCP_d will be able to output some really nice ALFF and Reho derivatives, but I am not certain what can XCP_d do for task-fMRI.
I understand that the task contrasts can be entered as custom “confounds”, and my question is, will these confounds only be regressed out (denoising), or will the first level subject spacial maps be generated from these contrasts like the first level output from fsl glm?
My understanding of task-fMRI is very limited, I am not sure what is the purpose of regressing out task effects, and use the residual in a connectivity analysis in the context of task fMRI. Could anyone please suggest some papers regarding this so that I can educate myself?
You can input the denoised timeseries from XCP_d into a task-based GLM model.
I believe they are only regressed out, and the maps for each confound are not saved out.
In terms of why you may want to regress out task activation, you can read more here: Task activations produce spurious but systematic inflation of task functional connectivity estimates - PMC. The basic idea is that, if you are looking at functional connectivity, you are probably interested in more about the interactions between areas as opposed to just the magnitudes of activations. Not controlling for the magnitudes of activations biases connectivity results.