JISCMail - FSL Archives

Did you try running FIX with the training file for that kind of data (i.e. Standard.RData)? I don’t personally have much experience with low spatial and temporal resolution data that is also very short. You could also try ICA-AROMA which might be appropriate for that kind of data.

Peace,

Matt.

From: FSL - FMRIB's Software Library <[log in to unmask]> on behalf of Noelle Dalin <[log in to unmask]>
Reply-To: FSL - FMRIB's Software Library <[log in to unmask]>
Date: Monday, November 28, 2016 at 1:38 PM
To: <[log in to unmask]>
Subject: Re: [FSL] Motion Scrubbing for rsFMRI

Hi Matt,

Thank you for the response! My fMRI data is from a 3T MRI at 3x3x3mm voxels, a TR of 3 seconds and only 120 volumes (approx. 6 minutes of data).

I haven’t been using FIX because our cohert is small with currently only approximately 60 participants , half of which likely have too much motion to even be included in the first place. I had intended to use FIX but, I don’t know if it worth while to train it for such a small sample. Instead I have been manually denoising using fsl_regfilt to remove noise components.

My reason for wanting to motion scrub is to improve the quality of the usable data especially since it is at rest.

Any suggestions or do you still think that I should skip it?

Thank you for your time!

Noelle Dalin

From: "[log in to unmask]" <[log in to unmask]> on behalf of Matt Glasser <[log in to unmask]>
Reply-To: "[log in to unmask]" <[log in to unmask]>
Date: Monday, November 28, 2016 at 11:25 AM
To: "[log in to unmask]" <[log in to unmask]>
Subject: Re: [FSL] Motion Scrubbing for rsFMRI

What kind of fMRI data is this (resolution, TR)? With HCP-Style fMRI data, you can rely on ICA+FIX to remove spatially localized noise (e.g. from subject head motion) and your use of the multi-variate dual regression approach will attenuate the effects of global physiological noise. So I wouldn’t worry about scrubbing.

Peace,

Matt.

From: FSL - FMRIB's Software Library <[log in to unmask]> on behalf of Noelle Dalin <[log in to unmask]>
Reply-To: FSL - FMRIB's Software Library <[log in to unmask]>
Date: Monday, November 28, 2016 at 1:16 PM
To: <[log in to unmask]>
Subject: [FSL] Motion Scrubbing for rsFMRI

Hello all!

I am working on motion scrubbing for resting-state functional connectivity data in a pediatric cohort with high motion.
I want to perform single session ICA to denoise and then group level ICA, Dual regression, and PALM after. From what I understand, my best option may be to use confound regression to remove high motion volumes(and those around them) rather than to actually delete volumes and replace them with mean (fsl_regfilt).

My command to get my confound file is:

fsl_motion_outliers -i PT#Letter_FC.nii.gz -o PT#Letter_fd_confound.txt -s PT#Letter_fd_metric.txt -p PT#Letter_fd_pic –-fd –v

I would then plug this confound file into FEAT>Stats tab>Add additional confound EVs and run feat with the melodic ICA data exploration (?- this is what I am trying out but I don’t know if it will work like I want it to as it is just running now. If you have a better method, please share. I am still learning)

Is there any way for me to have more control in terms of what volumes are regressed out (based on my own threshold for motion (.3000mm) and the removal of one volume before and two after) after rather than just relying on the confound file outputted from my Fsl_motion_outliers command? Is it better for me to use fsl_regfilt instead and identify the volumes that I want to remove based on my fd_metrics.txt?

Please advise.

Thank you for your time!

Noelle Dalin