NITRC: NITRC Community: open-discussion

open-discussion

28951 Subscribers

open-discussion > Connectivity question

Showing 1-5 of 5 posts

Connectivity question

Hello expert,

I have a question about fMRI connectivity analysis.

I’m trying to see if connectivity differs between groups of patient and controls, in a set of ROI.

Say I'd use correlation

My issue is that the fMRI timeseries are much longer for the control than patients, about 2 times longer, so more timepoints (~480 vs ~250); it's a feature of the task they were doing, the patients got fatigued much earlier. So, the correlation for the controls would be computed with more timepoints than the correlation of the patients.

-1-

So, my question is since the number of timepoints differ substantially between the groups, is this a problem for group analysis? I would compare the correlation between groups with a t-test for example.

-2-

If this an issue, is there a way out? Maybe up-sampling the patient time series or some other methods?

thansk a lot

patrick

RE: Connectivity question

Patrick,

I mainly work with resting state fMRI, but I have a suggestion for addressing your issue. You can always use fslsplit to split out the time series into individual volumes and then use fslmerge to merge the volumes you want back into one file for analysis. You could script this so it manually removes the extra volumes from the end of the time series in your patients.

Hope this helps,

Anna

RE: Connectivity question

Anna,

Thanks for responding!

I don't want to split the timeseries, I want to use all the timepoints (volumes) available.
But, this creates the issue of having many more timepoints for a group vs the other and so the correlations will be calculated with different number of timepoints.

Just to be sure I'm cler, the correlation between ROI for the controls would be based on 480 timepoints while the correlation for the patients would be based on 240 timepoints. Then I'd compare between groups those correlations.

Is this a problem that these correlations are based on different timeseries length?

best

patrick

RE: Connectivity question

Different numbers of timepoints can matter because the connectivity estimates from longer series are usually more stable. I would not fix that by upsampling. A better approach is to compute connectivity from matched portions of the task across all subjects (for example the same blocks or the first common segment), convert correlations to Fisher z, and then compare groups at the second level. You can include the number of usable frames as a covariate only as a secondary sensitivity analysis, but not as the main solution if scan length is essentially confounded with group.

RE: Connectivity question

ok thanks ,

I was afraid this would be the case!!

I did chopped the timeseries ans selected specific blocks (frames); but now I have less timepoints and the data is not continuous anymore, so it's harder to do causality analysis, say like Granger or some other time-varying FC