Positive not negative correlation maps in REST

Dear REST developers and experts
To the first, I really appreciate for providing a wonderful free software.
I've also attached the following query to the SPM mailing list.
I'm analysing seed-based (hippocampus) resting-state functional connectivity between patient and control groups by using spm8 and especially REST program.
The resulting zFC maps of include both positive and negative correlations.
However, negative correlation (or anti-correlation) seems pretty difficult to interpret the results, and moreover, increased negative correlation in patient groups results in significant positive clusters in 'control_to_patient' 2nd-level analysis using two-sample t-test, which makes me more confusing.
Therefore, I wish to remove negative correlation maps from the individual zFC map and just want to compare positive correlation maps between the groups (and I found some articles used this methods).
How can I do this?
Any hints would be appreciated.
Thanks in advance for your kindness.


Dear Kim:
I think that you want to remove the negative result of zFC maps, am I correct?
If so, I suggest you to run the following commands on MATLAB:
 X = rest_readfile('the path of your result'); % This command will help you to get your zFC map (named "X") as a 3-D matrix in MATLAB.
 X(find(X<0))==0; % Set the negative zFC results to zero.

Then you can use "rest_writefile ()" (details see "help") to save your new zFC map.

These three commands will help you make an individual zFC map done.  I hope it can help.


Zang Zhen-Xiang

Dear Dr. Zang

Thanks for the comment.
I faced another probelm as for the script X = rest_readfile('the path of your result')
What do you mean by 'the path of your result' ?
Is it the result directory where zFCmaps are in, or the individual file name (e..g, zFCMap_01.hdr, zFCMap_01.img)?
I have tried both, but in vain, resulting in error message.
So I want to know what exactly should I insert in the blanks.
Sorry for bothering you with my silly question.

Once again, thank you in advance.


Dear Kim,
The path of your zFC results is the directory of the result maps (.img/.hdr).
You have to make sure that REST is on your computer, firstly. Otherwise, the
function "rest_readfile" will be an undefined function.
For example:
X=rest_readfile('E\tmp\subject001.img') is available

Hope it can help
Zang Zhen-Xiang

Dear Zang and Yufeng

Thanks for the valuable comments.
Now, I can succeed in first and second steps you recommended.
However, with regard to the final (3rd command) step you wrote to me, I cannot understand what the 5 variables in the blanks of 'rest_writefile(data, filename, isize, vsize, dtype)' stand for? Which variables should I insert in the blanks?
I'm a begginer in this field of resting data analysis and earnestly hope that you kindly know me the right way.
Pretty sorry for bothering you again.
Many thanks.


Dear Dr. Kim,

Firstly, I do not recommend to remove such voxels. Some results are not easy to interpret. That's quite common.

Secondly, it should be cautious when perfoming group-level ststisitcal analysis if you want to remove a voxel showing negative correlation in an individual Z-map.  If you want to remove a voxel at individual level, you may have to remove this voxel from all participants because it would be problematic if the value for this voxel is zero for some of the participants. This voxel will show certainly 'no significant difference' between groups. But the actual result might be a "significant differrence", although difficult to explain.


Negative correlation usually appeared after global signal removal. However, this is still a controversial issue, please see more details in:
Murphy K, Birn RM, Handwerker DA, Jones TB, Bandettini PA (2009) The impact of
global signal regression on resting state correlations: Are anti-correlated networks
introduced? Neuroimage 44:893–905.

Fox MD, Zhang D, Snyder AZ, Raichle ME (2009) The global signal and observed
anticorrelated resting state brain networks. J Neurophysiol 101:3270–3283.

I agree with Dr. Zang's advice that you need to create a positive mask for all the group if you really want to remove the negative correlation.

You can use the GUI in REST to do this:
REST->Utilities->REST Image Calculator

1. First, create the positive mask for each individual:
Add all the participants to group 1, and use this expression: g1>0
You will get a new directory of images.
2. Sum the masks of each individual:
Add the output directory of the previous step: and use this expression: sum(g1)
You will get one result image.
3. Threshold on the sum image:
Add the result image as an individual image, and use this expression (if you have 81 participants): i1>81
Then you can get the mask. And you can use this mask in Statistical Analysis.

Please find more details about REST Image Calculator in http://www.restfmri.net/forum/Course
you can find it can help you to realize lots of ideas.

Dear Prof. YAN Chao-Gan and Zang

Thanks for your kindness.
I will try what you have indicated.
By the way, one expert in spm lists suggested the use of spm imcalc function, ie, 'i1.*(i1>0)' to remove negative correlation maps and to keep only voxels in which r>0.
Is the above-mentioned spm function is same to what you have suggested?
After having finished in making individual zFC maps with only positive correlation, I have performed 2nd level group analysis between patients and control subjects by using REST statistical analysis of two-sample t test, without an error. The results seemed similar to those of original zMC map analysis(containg both positive and negative maps) but the statistical significance was much lowered.
I was pretty disappointed and cannot understand how this results come out.
Thank you in advance for your kind comment,


If you just want to remove the negative correlation for each participant, then use spm 'i1.*(i1>0)' is OK.

The negative T values in two-sample t-test image is another issue.
If you have two groups, then positive T value means group 1 is higher than group 2, while negative T value means group 1 is lower than group 2.

I'd really appreciated your kindness.
Your comments are of much help to me.