help

help >

**Low correlation between upper and lower triangle FC matrix**Nov 12, 2021 09:11 PM | Selma Lugtmeijer -

*Brock University*Low correlation between upper and lower triangle FC matrix

Hi,

I read in a previous post an explanation why the FC between ROI 1 (seed) and ROI 2 (target) is different from ROI 2 (seed) and ROI 1 (target). However, in my matrix the correlation between the upper and lower triangle is .2, that seems very low, is that possible or is it more likely that I did something wrong?

What I did:

1) first level GLM for every subject

2) extract time series for all ROIs using VOI extract in SPM for every subject. Adjust for F-contrast and contrast of interest is the same F-contrast, no threshold

3) run the PPPI script for all ROIs for every subject

4) use the con image of the contrast of interest and get a mean value for all target ROIs for every subject using this function:

function ROI_data = Extract_ROI_Data_ppi2(ROI, Contrast)

Y = spm_read_vols(spm_vol(ROI),1);

indx = find(Y>0);

[x,y,z] = ind2sub(size(Y),indx);

XYZ = [x y z]';

ROI_data = nanmean(spm_get_data(Contrast, XYZ),2)

end

write that output in a 3d matrix (seed,target,subject)

5) get the mean values over subjects to make a connectivity plot

6) calculate the correlation with this code:

fc_l = [];fc_u = []; %lower and upper half of the diagonal of the matrix

for i=1:99 %100 ROIs

fc_l_i = FC_mean((i+1):100,i);

fc_l = [fc_l;fc_l_i]; %columns stacked, left lower half of the matrix

fc_u_i = FC_mean(i,(i+1):100);

fc_u = [fc_u,fc_u_i]; %rows next to each other, right upper half of the matrix

end

fc_u = fc_u.';

scatter(fc_u,fc_l)

[rho,pval]=corr(fc_u,fc_l)

Any thoughts on this are highly appreciated!

Selma

I read in a previous post an explanation why the FC between ROI 1 (seed) and ROI 2 (target) is different from ROI 2 (seed) and ROI 1 (target). However, in my matrix the correlation between the upper and lower triangle is .2, that seems very low, is that possible or is it more likely that I did something wrong?

What I did:

1) first level GLM for every subject

2) extract time series for all ROIs using VOI extract in SPM for every subject. Adjust for F-contrast and contrast of interest is the same F-contrast, no threshold

3) run the PPPI script for all ROIs for every subject

4) use the con image of the contrast of interest and get a mean value for all target ROIs for every subject using this function:

function ROI_data = Extract_ROI_Data_ppi2(ROI, Contrast)

Y = spm_read_vols(spm_vol(ROI),1);

indx = find(Y>0);

[x,y,z] = ind2sub(size(Y),indx);

XYZ = [x y z]';

ROI_data = nanmean(spm_get_data(Contrast, XYZ),2)

end

write that output in a 3d matrix (seed,target,subject)

5) get the mean values over subjects to make a connectivity plot

6) calculate the correlation with this code:

fc_l = [];fc_u = []; %lower and upper half of the diagonal of the matrix

for i=1:99 %100 ROIs

fc_l_i = FC_mean((i+1):100,i);

fc_l = [fc_l;fc_l_i]; %columns stacked, left lower half of the matrix

fc_u_i = FC_mean(i,(i+1):100);

fc_u = [fc_u,fc_u_i]; %rows next to each other, right upper half of the matrix

end

fc_u = fc_u.';

scatter(fc_u,fc_l)

[rho,pval]=corr(fc_u,fc_l)

Any thoughts on this are highly appreciated!

Selma