Iąd recommend using hcp-style fast TR and high resolution datasets with
fairly long run lengths (e.g. 15mins). ICA+FIX seems to do a pretty good
job with these datasets (as far as properly splitting signal and noise,
even when there is quite a bit of motion).
Peace,
Matt.
On 1/13/16, 1:10 PM, "FSL - FMRIB's Software Library on behalf of Colin
Hawco" <[log in to unmask] on behalf of [log in to unmask]> wrote:
>My experience with manual ICA correction(which I have been doing a lot)
>and Fix (which I have used only a bit) is that fix is not necessarily a
>good choice for high motion. A couple of large motion spikes can
>contaminate many ICs. You then get 'mixed' IC's which are good signal
>with a single or small number of large artifact spikes, which Fix will
>generally reject. As such, Fix will potentially remove a lot of true
>signal from the data, not just noise.
>
>There is another worry here, which is that if one group has high motion,
>many more components will be removed, which can change the character of
>the data and the connectivity patterns as a side effect of data
>filtering, not due to biology.
>
>Granted the same argument might be mad if you regress motion, as higher
>motion parameters may result in more data regressed out, and I really
>don't think motion paramters really properly capture motion artifacts. I
>have seen several participants with a large amount of ICA data which
>looks like motion spikes (to the point that I have to exclude them as all
>ICs get rejected), but the motion parameters don't look too bad, with max
>translations < 1mm.
>
>I believe, and have been told, that ICA is usually enough to regress CSF
>and WM signals, as they are usually components. However I have seen
>papers regress CSF and WM after ICA, and I don't think this would be a
>major issue.
>
>Motion is a tough nut to crack. One solution is to exclude TRs which
>appear to have motion artifact. I am not sure hwo the community feels
>about that at this time, but it was used in some older studies.
>
>Good luck!
>
>Colin Hawco, PhD
>Neuranalysis Consulting
>Neuroimaging analysis and consultation
>www.neuranalysis.com
>[log in to unmask]
>
>
>
>
>-----Original Message-----
>From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
>Behalf Of Sara De Simoni
>Sent: January-13-16 12:51 PM
>To: [log in to unmask]
>Subject: Re: [FSL] FIX - motion parameter correction
>
>Dear FSLers,
>
>As an additional question to that posted by Lucia -
>
>After using FIX (option -m) to clean your data, as well as avoiding
>regressing out the motion parameters following this, can you also avoid
>regressing out CSF and white matter signal? This is usually recommended
>but I am not sure if it is appropriate in combination with FIX.
>
>Also, it appears that FIX is a very good solution to dealing with noise
>in low-motion datasets but has it been shown to be an efficient clean up
>in the context of resting-state data where a lot of motion related noise
>is present?
>
>i.e. does it avoid the biases in functional connectivity that you find
>with high-motion data? I understand that this is quite a general question
>but I would be very grateful for any advice/info you may have!
>
>Thanks for your help!
>
>Sara
>
>Sara De Simoni
>Post-doctoral research associate
>Imperial College London
|
