Re: [RFMRI] Issue with Nuisance Covariates Regression_additional question

Submitted by phil1022 on

YAN Chao-Gan

Thu, 04/02/2015 - 01:29

Permalink

Hi Xiao-Song,
This is normal. If you don't do global signal regression, most of the correlations are positive. If you do global signal regression, half will be negative. You can read some articles about the global signal regression -- it's still a controversial. You can also read http://rfmri.org/GSRDiscussion to get some information.

Best,

Chao-Gan 

On Wed, Apr 1, 2015 at 1:50 PM, The R-fMRI Network <rfmri.org@gmail.com> wrote:
[To post a comment, please reply to rfmri.org@gmail.com ABOVE this line]

By Xiaosong He (phil1022)

Dear Chao-Gan,

So surprisingly, based on the FAQ, what I saw is normal! Thanks for
clarifying that otherwise I would never believe it. I used to do the
regression by SPM8/REST (originally by AFNI) to eliminate the head motion
and WM & CSF, none of my previous methods had ever given me that kind of
results before.

Anyway, actually the reason for me to look back into these regressed
images is because, I find some difference between the results given by
DPARSF and SPM8/REST. The main purpose here is to preprocess the rsfMRI
data, then exract the BOLD signal based on AAL template, for further
graphic theory related analysis. DPARSF can perfectly output the
correlation map for the 116 ROIs of AAL. The thing is, I find that the
correlations looked too high than they should be. Let's say, normally,
among the 90 cerebral ROIs, you may have 50% correlations which are
negative. In specific, one of my dataset of over 50 subjects which was
processed by SPM8/REST, none of them have more than 45% correlations stayed
positive (the r value of the rest 55% are no more than 0). However, in my
recent dataset of 35 subjects which was processed by DPARSF, none of them
have less than 70% correlations stay positive (in a few of them that
proportion reached 90%). I mean, this is highly impossible, right? At least
there should be the anti-correlation networks existing that the
correlations between them should stay negative all the way.

Do you have any idea where I may make some mistake to cause this
result? I looked into the snapshot of DPARSFA on the rfmri.org website, my
setting is almost identical except I did not reorientate the functional
image which can barely affect the result.

Thank you very much!

With kind regards.

Sincerely,
Xiao-Song


Online version of this post: http://www.rfmri.org/content/re-rfmri-issue-nuisance-covariates-regressionadditional-question


Many a little makes a mickle -- your kind contributions shall make our efforts not perish from the earth. Please help The R-fMRI Network at http://rfmri.org/#overlay=HelpUs
To manage subscriptions, please visit: http://rnet.co/mailman/listinfo/rfmri.org_rnet.co
Mail comment ID: http://www.rfmri.org/mailcomment/redirect/%3C31.1994.0.1427910599.3e66484b99db759378572f52ec944f95%40www.rfmri.org%3E

Dear Chao-Gan,
    You know what? I really want work with you. There are so many things I can learn from you. Thank you very much for your patience and I am sincerely appreciating all your help.

    With kind regards.

    Sincerely,

    Xiao-Song

On Wed, Apr 1, 2015 at 9:30 PM, The R-fMRI Network <rfmri.org@gmail.com> wrote:
[To post a comment, please reply to rfmri.org@gmail.com ABOVE this line]

Commented by YAN Chao-Gan (YAN Chao-Gan)
Hi Xiao-Song,
This is normal. If you don't do global signal regression, most of the correlations are positive. If you do global signal regression, half will be negative. You can read some articles about the global signal regression -- it's still a controversial. You can also read http://rfmri.org/GSRDiscussion to get some information.

Best,

Chao-Gan 

On Wed, Apr 1, 2015 at 1:50 PM, The R-fMRI Network <rfmri.org@gmail.com> wrote:
[To post a comment, please reply to rfmri.org@gmail.com ABOVE this line]

By Xiaosong He (phil1022)

Dear Chao-Gan,

So surprisingly, based on the FAQ, what I saw is normal! Thanks for
clarifying that otherwise I would never believe it. I used to do the
regression by SPM8/REST (originally by AFNI) to eliminate the head motion
and WM & CSF, none of my previous methods had ever given me that kind of
results before.

Anyway, actually the reason for me to look back into these regressed
images is because, I find some difference between the results given by
DPARSF and SPM8/REST. The main purpose here is to preprocess the rsfMRI
data, then exract the BOLD signal based on AAL template, for further
graphic theory related analysis. DPARSF can perfectly output the
correlation map for the 116 ROIs of AAL. The thing is, I find that the
correlations looked too high than they should be. Let's say, normally,
among the 90 cerebral ROIs, you may have 50% correlations which are
negative. In specific, one of my dataset of over 50 subjects which was
processed by SPM8/REST, none of them have more than 45% correlations stayed
positive (the r value of the rest 55% are no more than 0). However, in my
recent dataset of 35 subjects which was processed by DPARSF, none of them
have less than 70% correlations stay positive (in a few of them that
proportion reached 90%). I mean, this is highly impossible, right? At least
there should be the anti-correlation networks existing that the
correlations between them should stay negative all the way.

Do you have any idea where I may make some mistake to cause this
result? I looked into the snapshot of DPARSFA on the rfmri.org website, my
setting is almost identical except I did not reorientate the functional
image which can barely affect the result.

Thank you very much!

With kind regards.

Sincerely,
Xiao-Song


Online version of this post: http://www.rfmri.org/content/re-rfmri-issue-nuisance-covariates-regressionadditional-question


Many a little makes a mickle -- your kind contributions shall make our efforts not perish from the earth. Please help The R-fMRI Network at http://rfmri.org/#overlay=HelpUs
To manage subscriptions, please visit: http://rnet.co/mailman/listinfo/rfmri.org_rnet.co
Mail comment ID: http://www.rfmri.org/mailcomment/redirect/%3C31.1994.0.1427910599.3e66484b99db759378572f52ec944f95%40www.rfmri.org%3E


Online version of this post: http://www.rfmri.org/comment/3622#comment-3622


Many a little makes a mickle -- your kind contributions shall make our efforts not perish from the earth. Please help The R-fMRI Network at http://rfmri.org/#overlay=HelpUs
To manage subscriptions, please visit: http://rnet.co/mailman/listinfo/rfmri.org_rnet.co
Mail comment ID: http://www.rfmri.org/mailcomment/redirect/%3C31.1994.3622.1427938199.0b7af0c61a9da31f94fd8f4169edf8b3%40www.rfmri.org%3E

Dear Chao-Gan, So surprisingly, based on the FAQ, what I saw is normal! Thanks for clarifying that otherwise I would never believe it. I used to do the regression by SPM8/REST (originally by AFNI) to eliminate the head motion and WM & CSF, none of my previous methods had ever given me that kind of results before. Anyway, actually the reason for me to look back into these regressed images is because, I find some difference between the results given by DPARSF and SPM8/REST. The main purpose here is to preprocess the rsfMRI data, then exract the BOLD signal based on AAL template, for further graphic theory related analysis. DPARSF can perfectly output the correlation map for the 116 ROIs of AAL. The thing is, I find that the correlations looked too high than they should be. Let's say, normally, among the 90 cerebral ROIs, you may have 50% correlations which are negative. In specific, one of my dataset of over 50 subjects which was processed by SPM8/REST, none of them have more than 45% correlations stayed positive (the r value of the rest 55% are no more than 0). However, in my recent dataset of 35 subjects which was processed by DPARSF, none of them have less than 70% correlations stay positive (in a few of them that proportion reached 90%). I mean, this is highly impossible, right? At least there should be the anti-correlation networks existing that the correlations between them should stay negative all the way. Do you have any idea where I may make some mistake to cause this result? I looked into the snapshot of DPARSFA on the rfmri.org website, my setting is almost identical except I did not reorientate the functional image which can barely affect the result. Thank you very much! With kind regards. Sincerely, Xiao-Song On Wed, Apr 1, 2015 at 11:57 AM, The R-fMRI Network wrote: > [To post a comment, please reply to rfmri.org@gmail.com ABOVE this > line] > > Commented by YAN Chao-Gan (YAN Chao-Gan) > > Hi Xiao-Song, > > You can have a look at http://rfmri.org/FAQ #4: > http://www.rfmri.org/node/1207#comment-1906 > > Best, > > Chao-Gan > > Online version of this post: > http://www.rfmri.org/comment/3616#comment-3616 > > > Many a little makes a mickle -- your kind contributions shall make our > efforts not perish from the earth. Please help The R-fMRI Network at > http://rfmri.org/#overlay=HelpUs > To manage subscriptions, please visit: > http://rnet.co/mailman/listinfo/rfmri.org_rnet.co > Mail comment ID: > http://www.rfmri.org/mailcomment/redirect/%3C31.1991.3616.1427903826.2c6976688cdaa684e3113ded5a4e11f6%40www.rfmri.org%3E >