No, that would not be valid (unless you are going to again include the confound file as part of fitting your seed based connectivity). Whether you get the residuals file using fsl_regfilt or fsl_glm is irrelevant. The confound matrix returned by fsl_motion_outliers
is not really intended to be used to a "residualize" a time series. Rather, it is intended to be used in the context of an overall GLM in which you want to ignore the impact of certain frames on the model fitting.
cheers,
-MH
-----------------------------------------------------------
Conte Center for the Neuroscience of Mental Disorders
Washington University School of Medicine
Department of Psychiatry, Box 8134
660 South Euclid Ave.
Tel: 314-747-6173
Dear MH
I have one question about the two step analysis approach in seed based functional connectivity study. I have a dataset which has already regress out the effect of motion parameters, WM and CSF signal. Currently I want to regress out the effect of the bad
timepoints which are identified using fsl_motion_outliers for these dataset. After reading the posts of "fsl_motion_outliers question", I realize such kind analysis approach may have some problem but the discussion is all about using the command of "fsl_regflit".
Is it valid If I use fsl_glm to regress out the effect of bad timepoints and take the residual part to further perform seed based functional connectivity analysis ?
Best
Paul
Date: Fri, 15 Aug 2014 16:41:42 +0000
From:
[log in to unmask]
Subject: Re: [FSL] fsl_motion_outliers - resting-state data
To:
[log in to unmask]
Hi,
This question should have a FAQ on the FSL Wiki, because it comes up repeatedly. See the "fsl_motion_outliers question" thread from end of May 2014 for a recent set of posts on this issue.
The key point I what to raise here is that, depending on exactly how you intend to analyze the data, you can NOT simply regress out the confound matrix returned by fsl_motion_outliers and then proceed to analyze the ensuing time series of "residuals".
The frames identified by the confound matrix will effectively be replaced by the mean signal for those time points, and one needs to be aware of that, and take it into account in any downstream analysis (i.e., by making sure that you also effectively exclude
those frames in any downstream analysis as well). This was the point of the thread from the end of May.
cheers,
-MH
-----------------------------------------------------------
Conte Center for the Neuroscience of Mental Disorders
Washington University School of Medicine
Department of Psychiatry, Box 8134
660 South Euclid Ave.Tel: 314-747-6173
Ryan - I did this recently for my resting state data so I can tell you what worked for me.
First: It was suggested by the FSL gurus that this would be done AFTER removing junk components (if you're doing ICA) using fsl_regfilt.
Second: To apply the motion.txt file output of fsl_motion_outliers to your regfilted data you use the stats only tab in the FEAT gui and insert those .txt files into the "extra EVs" portion of the stats tab. Make sure they're in the same order as your
input data. You can get around having to do any modeling business by just saving your design file prior to clicking "go."
The resulting output will be a new "filename.feat" directory within each subect's primary feat directory - at least it was for me b/c I put it there. You can then go on to do group analyses with that output.
Hopefully that makes sense.
Cheers
The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended
recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone
or return mail.
