How to deal with different results with and without global signal regression

Submitted by yaosir on

Dear experts,

I have collected the data of two groups of participants (patients - controls). Firstly, I analyzed the data without global signal regression, and there were some clusters remaining (all Patients > controls). Then, I analyzed the data with global signal regression, but there were only a remaining cluster (Patients < controls). I am really puzzeled about the very different results with and withough GSR (number of cluters and the valence of the clusters). And the questions such as GSR and head motion made me feel somewhat depressed since I entered the field of resting-state fMRI. I am really appreciated if you tell me some ways to deal with different results with and without GSR in this case.

Thanks in advance,

Vincent

YAN Chao-Gan

Sun, 09/07/2014 - 11:03

Hi Vincent,

Yes, we all feel depressed as this field is still suffering so many methodological issues -- and even for preprocessing, there are so many debating topics!!! However, this makes us have a job, which makes us can feed ourselves, right? ;)

For paper reviewing wise, if you go ahead without Global Signal Regression (GSR), almost all the reviewers will not give you a hard time on this issue.

But if you go ahead with GSR, then most reviewers will give you a really hard time.

Thus, I will recommend you go ahead without GSR to reduce your headache. Although I do think GSR has its value in my own perspective, and I may put them into supplementary results.

Hope this helps.

Best,

Chao-Gan

Dear Dr. Yan,

Thank you for your encouragement and helpful advice!

Luckily, the results are much better without GSR in this case. However, just in case, I wonder how to respond if reviewer asks me the reasons that I did not do GSR? Is it OK to cite some papers that suggest not to do GSR (e.g., Murphy et al., 2009)?

Best,

Vincent

Hi Vincent,

Yes, you can cite that one. You can also cite Saad's work:

Saad, Z.S., Gotts, S.J., Murphy, K., Chen, G., Jo, H.J., Martin, A., Cox, R.W., 2012. Trouble at rest: how correlation patterns and group differences become distorted after global signal regression. Brain Connect 2, 25-32.

Saad, Z., Reynolds, R.C., Jo, H.J., Gotts, S.J., Chen, G., Martin, A., Cox, R., 2013. Correcting Brain-Wide Correlation Differences in Resting-State FMRI. Brain Connect 3, 339-352.

Now I am seriously kidding: please show me the proof that you have participated the social media event "The Wrath of TRN" to have your next question answered. ;)

Best,

Chao-Gan

Dear Vincent & Dear Yan

Please let me share  some background (be cautious with a Rookie :P)

About GSR, even with the opposite, when GSR it's your unique option you can check this paper Chai et al., 2011 Neuroimage. http://www.sciencedirect.com/science/article/pii/S1053811911009657  

Because till I know Murphy, 2009 said that you must take care your results when there are anti-correlated networks. 

Murphy = "These results call into question the interpretation of negatively correlated regions in the brain when using global signal regression as an initial processing step."

Chai = "Our results suggest that anticorrelations observed in resting-state connectivity are not an artifact introduced by global signal regression and might have biological origins"

In my case, my data looks better with GSR, personally I preffer it. 

So, this is a House Of Cards. 

Thank you all,

EDIT: I've realized that my answer it could be confusing. Chai et al., used the aCompCor method to defend their job and say the question above. It's important to know that you can find and a version of CompCor in DPARSFA/dpabi.

Hi Susan,
How are you?

I am on the bus to Boston now, and I am happy that I will see you soon!

Actually, here is a post by Joan, he read your Neuroimage paper very carefully! Would you mind to reply him directly? I don't want to speak for you as I am afraid that I may mis-interpret your work. In Chinese, we call that 越俎代庖。

You can reply this email directly to rfmri.org@gmail.com if you have an account at RFMRI.ORG already. If not, you can open one if you want, then all the community of The R-fMRI Network can learn from you.

Thanks and see you soon!

Best,

Chao-Gan

On Sun, Sep 7, 2014 at 11:29 AM, 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 Joan (Ham)

Dear Vincent & Dear Yan

Please let me share  some background (be cautious with a Rookie :P)

About GSR, even with the opposite, when GSR it's your unique option you can check this paper Chai et al., 2011 Neuroimage. http://www.sciencedirect.com/science/article/pii/S1053811911009657  

Because till I know Murphy, 2009 said that you must take care your results when there are anti-correlated networks. 

Murphy = "These results call into question the interpretation of negatively correlated regions in the brain when using global signal regression as an initial processing step."

Chai = "Our results suggest that anticorrelations observed in resting-state connectivity are not an artifact introduced by global signal regression and might have biological origins"

In my case, my data looks better with GSR, personally I preffer it. 

So, this is a House Of Cards. 

Thank you all,

EDIT: I've realized that my answer it could be confusing. Chai et al., used the aCompCor method to defend their job and say the question above. It's important to know that you can find and a version of CompCor in DPARSFA/dpabi.


Online version of this post: http://rfmri.org/comment/3238#comment-3238


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://rfmri.org/mailcomment/redirect/%3C31.1815.3238.1410103754.91137d9f3e1ba6e4e68e51469033289f%40www.rfmri.org%3E

Hi Ziad,
How are you? I know your are going to present "Global Correlations: What You Don’t Know Will Hurt You" in about an hour: http://www.martinos.org/brainconnectivity/pre-conference-workshop/

But why not put your comment here also? People here are also eager to be aware that issue as well!

You can reply this email directly to rfmri.org@gmail.com if you have an account at RFMRI.ORG already. If not, you can open one if you want, then all the community of The R-fMRI Network can learn from you.

Thanks and and looking forward to your talk! (I am sitting at the first row, and hope you can see me. ;) )

Best,

Chao-Gan

On Mon, Sep 8, 2014 at 4:01 PM, Susan Gabrieli <swg@mit.edu> wrote:

Dear Chao-Gan

So great to hear from you!

I will read this asap – I’m slammed putting out conference related fires – which I hope are resolved before you arrive! ;)

J

Cheers,

Susan

 

 

From: YAN Chao-Gan [mailto:ycg.yan@gmail.com]
Sent: Monday, September 08, 2014 3:52 PM
To: The R-fMRI Network
Cc: Susan Gabrieli
Subject: Re: [RFMRI] How to deal with different results with and without global signal regression

 

Hi Susan,

 

How are you?

 

I am on the bus to Boston now, and I am happy that I will see you soon!

 

Actually, here is a post by Joan, he read your Neuroimage paper very carefully! Would you mind to reply him directly? I don't want to speak for you as I am afraid that I may mis-interpret your work. In Chinese, we call that 越俎代庖。

 

You can reply this email directly to rfmri.org@gmail.com if you have an account at RFMRI.ORG already. If not, you can open one if you want, then all the community of The R-fMRI Network can learn from you.

 

Thanks and see you soon!

 

Best,

 

Chao-Gan

 

On Sun, Sep 7, 2014 at 11:29 AM, 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 Joan (Ham)

Dear Vincent & Dear Yan

Please let me share  some background (be cautious with a Rookie :P)

About GSR, even with the opposite, when GSR it's your unique option you can check this paper Chai et al., 2011 Neuroimage. http://www.sciencedirect.com/science/article/pii/S1053811911009657  

Because till I know Murphy, 2009 said that you must take care your results when there are anti-correlated networks. 

Murphy = "These results call into question the interpretation of negatively correlated regions in the brain when using global signal regression as an initial processing step."

Chai = "Our results suggest that anticorrelations observed in resting-state connectivity are not an artifact introduced by global signal regression and might have biological origins"

In my case, my data looks better with GSR, personally I preffer it. 

So, this is a House Of Cards. 

Thank you all,

EDIT: I've realized that my answer it could be confusing. Chai et al., used the aCompCor method to defend their job and say the question above. It's important to know that you can find and a version of CompCor in DPARSFA/dpabi.


Online version of this post: http://rfmri.org/comment/3238#comment-3238

 

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://rfmri.org/mailcomment/redirect/%3C31.1815.3238.1410103754.91137d9f3e1ba6e4e68e51469033289f%40www.rfmri.org%3E

 


Dear Dr. Murphy,
Hope all is well!

Actually, here is a post by Joan, he read your Neuroimage paper very carefully! Would you mind to reply him directly?

I would like to invite related scholars, such as Drs. Susan Whitfield-Gabrieli, Ziad Saad... into this Global Signal Regression thread. Here I would like to invite you to put your comments on this thread. 

You can reply this email directly to rfmri.org@gmail.com if you have an account at RFMRI.ORG already. If not, you can open one if you want, then all the community of The R-fMRI Network can learn from you.

Thanks and see you soon (I know you have a poster here at the Fourth Biennial Conference on Resting State / Brain Connectivity and looking forward to seeing you here)!

Best,

Chao-Gan


On Tue, Sep 9, 2014 at 9:57 AM, YAN Chao-Gan <ycg.yan@gmail.com> wrote:
Hi Ziad,

How are you? I know your are going to present "Global Correlations: What You Don’t Know Will Hurt You" in about an hour: http://www.martinos.org/brainconnectivity/pre-conference-workshop/

But why not put your comment here also? People here are also eager to be aware that issue as well!

You can reply this email directly to rfmri.org@gmail.com if you have an account at RFMRI.ORG already. If not, you can open one if you want, then all the community of The R-fMRI Network can learn from you.

Thanks and and looking forward to your talk! (I am sitting at the first row, and hope you can see me. ;) )

Best,

Chao-Gan

On Mon, Sep 8, 2014 at 4:01 PM, Susan Gabrieli <swg@mit.edu> wrote:

Dear Chao-Gan

So great to hear from you!

I will read this asap – I’m slammed putting out conference related fires – which I hope are resolved before you arrive! ;)

J

Cheers,

Susan

 

 

From: YAN Chao-Gan [mailto:ycg.yan@gmail.com]
Sent: Monday, September 08, 2014 3:52 PM
To: The R-fMRI Network
Cc: Susan Gabrieli
Subject: Re: [RFMRI] How to deal with different results with and without global signal regression

 

Hi Susan,

 

How are you?

 

I am on the bus to Boston now, and I am happy that I will see you soon!

 

Actually, here is a post by Joan, he read your Neuroimage paper very carefully! Would you mind to reply him directly? I don't want to speak for you as I am afraid that I may mis-interpret your work. In Chinese, we call that 越俎代庖。

 

You can reply this email directly to rfmri.org@gmail.com if you have an account at RFMRI.ORG already. If not, you can open one if you want, then all the community of The R-fMRI Network can learn from you.

 

Thanks and see you soon!

 

Best,

 

Chao-Gan

 

On Sun, Sep 7, 2014 at 11:29 AM, 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 Joan (Ham)

Dear Vincent & Dear Yan

Please let me share  some background (be cautious with a Rookie :P)

About GSR, even with the opposite, when GSR it's your unique option you can check this paper Chai et al., 2011 Neuroimage. http://www.sciencedirect.com/science/article/pii/S1053811911009657  

Because till I know Murphy, 2009 said that you must take care your results when there are anti-correlated networks. 

Murphy = "These results call into question the interpretation of negatively correlated regions in the brain when using global signal regression as an initial processing step."

Chai = "Our results suggest that anticorrelations observed in resting-state connectivity are not an artifact introduced by global signal regression and might have biological origins"

In my case, my data looks better with GSR, personally I preffer it. 

So, this is a House Of Cards. 

Thank you all,

EDIT: I've realized that my answer it could be confusing. Chai et al., used the aCompCor method to defend their job and say the question above. It's important to know that you can find and a version of CompCor in DPARSFA/dpabi.


Online version of this post: http://rfmri.org/comment/3238#comment-3238

 

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://rfmri.org/mailcomment/redirect/%3C31.1815.3238.1410103754.91137d9f3e1ba6e4e68e51469033289f%40www.rfmri.org%3E