Hi Matt,

I'd actually already seen the preprint in bioxiv.org.  Given how strongly we 
like ICA-based solutions for things, I'm looking forward to being able to 
implement that approach in a standard way in our processing pipeline.  My 
problem was more immediate - I'd like, if possible, do produce data that fit 
the "vague consensus" that graph theory-using neuroimagers have tentatively 
reached over the past few years as the result of lots of work by Power, 
Satterthwaite and the rest of the bunch who've tackled the daunting problem of 
comparing various combinations of preprocessing.  I sit far closer to your 
position... I've always thought that ICA could be tweaked in ways that probably 
would do a better job of identifying and cleaning physiological and/or other 
types of structured noise.

But my issue is just more time-sensitive.  Sure, I could whip up a simple 
cross-correlation of our timeseries data WITHOUT regressing WM and CSF.  And 
what I have in mind is literally, just a teaser for some talks this weekend at 
a small, topically-focused conference.  And if I were a bit more savvy with HCP 
tools, I'd know if it were possible to re-create that volume-based file on my 
own.  I've already done the WM and CSF regression from the *.nii files on other 
GT-focused analyses using code I extracted from the RestingState.sh script.  (I 
know WM and CSF isn't saved in CIFTI).  My problem is simply I clobbered the 
corresponding volume-based *.nii for space-saving reasons, so I can't pull the 
relevant FS-labled WM and CSF values out.  I'm wondering if this can be done.  
If not, I won't waste any more thought on it.

Again, this is not crucial by any means... It's just if I'm going to do 
something that I already know is a bit sub-optimal, I'd like it to be "the 
least sub-optimal" I can possibly do in the timeframe I have available.  In 
this case, I just have a couple of days to assemble some nice preliminary 
results.

Best,
Mike


From: Glasser, Matthew [mailto:[email protected]]
Sent: Sunday, July 22, 2018 10:22 AM
To: Stevens, Michael; [email protected]
Subject: Re: [HCP-Users] WM and CSF signal extraction

This is an email from an external source. USE CAUTION opening attachments or 
links from unknown senders.

Hi Michael,

WM and CSF aren't saved in CIFTI.  Also I don't think that you gain much by 
regressing those out and might even lose some.  sICA+FIX will deal with 
movement effects but does not handle global physiological noise (which WM and 
CSF don't do a good job with either).  You can use temporal ICA to remove 
global physiological noise as will eventually be coming out in Neuroimage once 
they decide to make some correct proofs...

https://www.biorxiv.org/content/early/2018/03/25/193862<https://urldefense.proofpoint.com/v2/url?u=https-3A__www.biorxiv.org_content_early_2018_03_25_193862&d=DwMF-g&c=e_HtEeZEQXP5NUOb33qoTj0AVvRFBS9_rhBTQcfkWoA&r=golSXqwzwyESse15ixCMA3FkCV08C4TvBJ1P6uKO0nE&m=sdiQwdVLc3DEBFKYQfH4nhD128O9fuXHMQkHTosYA6c&s=5vMtwHSGwIE0-NtN4tkLq5DfBj0_SjOJkxT6HappiaY&e=>

Matt.

From: 
<[email protected]<mailto:[email protected]>>
 on behalf of "Stevens, Michael" 
<[email protected]<mailto:[email protected]>>
Date: Sunday, July 22, 2018 at 9:14 AM
To: "[email protected]<mailto:[email protected]>" 
<[email protected]<mailto:[email protected]>>
Subject: [HCP-Users] WM and CSF signal extraction

Hi folks,

I apologize in advance for asking a highly specific question for which I 
already know there's a better answer than what I'm asking.

Long story short -- I need to know if there's a way to reproduce the final 
output nifti volume file (in my pipeline case... the "clean" version that comes 
out of ICA-FIX) from the final CIFTI output.  To explain, for space saving 
reasons on our Linux storage array, we truncate files we don't think we need 
for CIFTI stats.  Because we never need that volumetric file for anything, 
we've killed it for a couple hundred datasets on one of our R01 projects.  But 
I probably need to quickly whip up some graph theory-style connectivity results 
on these data this coming week for a series of talks we're giving next weekend. 
 Because that volumetric file is gone in this dataset, I can't use the lines of 
code from the RestingState.sh script that produces the WM.txt and CSF.txt files 
so I can regress out those artifact signals in each timeseries (thanks to the 
prior posts on the HCP listserv for pointing me to that code).

So my question - Is there some useful combination of wb_command function calls 
that I can use to convert the final CIFTI product of the timeseries back into 
volume space (e.g., using the surface information that we were hopefully smart 
enough NOT to eradicate to save space)?  I've already successfully worked up 
code that finds the WM and CSF loci via the FreeSurferWMRegLut.txt and 
FreeSurferCSFRegLut.txt files to extract timeseries-specific WM and CSF values. 
 And again, I know the better way to do this probably would be to simply 
integrate a version of the RestingState.sh batch file into my processing 
pipeline and re-run things from scratch so ALL the data we need for this is 
produced.  But my short time-frame this week prevents that... Even on our 
cluster, it'd take all week or longer to re-process >300 datasets.  So if this 
is possible and someone can point me to the right commands/files, I'd 
appreciate it.  In the end, these would just be "teaser" slides of new data.  
But I hate showing anything... even preliminary results... when I know the 
methodology isn't as accurate as I can make it.

Thanks!
Mike


Michael C. Stevens, Ph.D.
Director, CNDLAB, Olin Neuropsychiatry Research Center
Director, Child & Adolescent Research, The Institute of Living
Adjunct Professor of Psychiatry, Yale University School of Medicine



This e-mail message, including any attachments, is for the sole use of the 
intended recipient(s) and may contain confidential and privileged information. 
Any unauthorized review, use, disclosure, or distribution is prohibited. If you 
are not the intended recipient, or an employee or agent responsible for 
delivering the message to the intended recipient, please contact the sender by 
reply e-mail and destroy all copies of the original message, including any 
attachments.

_______________________________________________
HCP-Users mailing list
[email protected]<mailto:[email protected]>
http://lists.humanconnectome.org/mailman/listinfo/hcp-users<https://urldefense.proofpoint.com/v2/url?u=http-3A__lists.humanconnectome.org_mailman_listinfo_hcp-2Dusers&d=DwMF-g&c=e_HtEeZEQXP5NUOb33qoTj0AVvRFBS9_rhBTQcfkWoA&r=golSXqwzwyESse15ixCMA3FkCV08C4TvBJ1P6uKO0nE&m=sdiQwdVLc3DEBFKYQfH4nhD128O9fuXHMQkHTosYA6c&s=52NzUImZA0cl1syw8DV-dZGw-AWUBD4N8ZqHT4PBZRc&e=>

________________________________
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.


Reminder: This e-mail and any attachments are subject to the current HHC email 
retention policies. Please save or store appropriately in accordance with 
policy.

_______________________________________________
HCP-Users mailing list
[email protected]
http://lists.humanconnectome.org/mailman/listinfo/hcp-users

Reply via email to