Hi Nia,
This topic is probably more relevant for the GIFT list, I replied
to your message there.
Best,
Vince
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On
Behalf Of Nia Goulden
Sent: Tuesday, December 10, 2013 2:45 AM
To: [log in to unmask]
Subject: [SPM] Combining DCM and ICA
Dear all
I have received some reviewer comments for a paper which I had submitted and
I would appreciate some advice on how to address the comments.
I have resting state data. I pre-process the data as usual, with slice
timing, realignment, normalisation using DARTEL and smoothing.
I then enter the smoothed data into a constrained ICA analysis using GIFT. I
use templates supplied with GIFT for the default mode network (DMN), the
salience network (SN) and the central executive network (CEN) in order to
obtain a component for each of these networks.
Having done this I have a time series for each component for each
participant. I enter these time series into a DCM model with three nodes,
one for each network/component. I specify full intrinsic connectivity
between the three nodes and nonlinear modulations in order to investigate
the relationship and switching between the networks.
The reviewers have stated that I should apply physiological noise regression
to the data. I did not collect physiological data so I am not able to do
this at the moment.
The reviewers have also stated that I should regress out motion from the
time series prior to entering the data into GIFT. I have tried doing this
with DPARSF and GIFT will not run when this is done, stating an empty
analysis. Is there another way to do this or to incorporate the motion
parameters into GIFT?
Lastly one reviewer questions why I do not use the templates for the
networks and take the average time series from the network. The reason why I
chose ICA was because I thought this would provide a more representative
time series for the network and I thought that taking an average or first
eigenvariate across a large region may be problematic. Does anyone have any
thoughts on this? Would this be a better approach?
Any advice, thoughts or comments appreciated
Many thanks
Nia
