Hi Joe,
On 12 Jul 2015, at 06:32, Jonathan Starke <[log in to unmask]> wrote:
> Dear Ludovica,
>
> thanks again for the assistance. i have some further queries about how to best proceed with the analysis i am interested in. for input date, i have separate stage 1 dual regression output (activation time series) for 3 groups of subjects. the groups are: a patient group (14 subjects) before and then after an intervention, and a control group (10 subjects). what i would like to do is determine the nature of between-network relationships for each group separately first. the group_maps i am using are the 10 network templates from the PNAS smith 2009 paper. i am happy with how to set up this initial part of the nets_examples script, and how to specify the analysis i want, e.g. full vs partial correlation.
> the script then asks for a glm. i have existing glms which i have used in other parts of my overall analysis (e.g. for randomise), but these specify btw-group comparisons (e.g. patients at baseline vs controls). since in the first step of FSLnets i am interested in the within-group situation, these existing glms would be innapropriate. i am then uncertain what form the glm being asked for should take? should it simply be the network matrix i am interested in, i.e. ten by ten networks? or the ten networks by number of subjects, prepared for each group?
The design matrix required by nets_examples will be used only by the function nets_glm to compare networks across subjects.
After you loaded the data and calculated the networks (nets_netmats), you’ll have one 10x10 matrix per subject.
I’m assuming you’re running FSLNets separately for each group, so if you simply want to do a within-group analysis you might just want to set a design matrix do a one-sample t-test for each group.
>
> as a second step, i would then like to compare network relationships found in step one differ between groups, specifically to look at any differences btw the patient group at baseline (before the intervention) and the control group, and also to compare the patient group before and after the intervention.
> my impression is that the nets_examples script can help me with the first step only, but that the second i will need to do in some other way. is this correct?
If you ran the dual regression on 3 groups separately, you’d have to combine the networks calculated from each group (output of nets_netmats) together in Matlab and then run the comparisons between groups setting the appropriate design matrix (e.g. patients at baseline vs controls etc…).
An alternative would be to re-run dual regression on all subjects together, so that when you load them with FSLNets you already have all your subjects combined.
However, you need to keep in mind that dual regression is run within a common mask derived from the specific group of subjects. Therefore if you run dual regression on the 3 groups separately or on all subjects together you’ll get slightly different results.
>
>
> so, for the second step to determine any differences in network relationships btw the different groups, what part of the FSLnets output should i use?
output of nets_netmats as input for nets_glm - see above
> and how to do this comparison, to determine the nature of the difference (e.g. DMN-executive control network becoming more or less correlated after the intervention) rather than simply the presence of a difference? should it be through calculating the presence and direction of the change in correlation co-efficients as apposed to say a t-test.
After the t-test, you can check the directionality of the significant changes by plotting the connection of interest with nets_boxplots (you can see examples of interpretation of the changes here http://fsl.fmrib.ox.ac.uk/fslcourse/lectures/practicals/rfMRIconnectivity/index.html)
>
> i hope i have explained my situation well.
> any advice/guidance would be greatly appreciated.
>
> thanks
> joe starke
> masters student
> university of cape town
>
Best,
Ludovica
—
Ludovica Griffanti, PhD
Analysis Postdoctoral Research Assistant
Oxford Centre for Functional MRI of the Brain (FMRIB)
Nuffield Department of Clinical Neurosciences, University of Oxford
John Radcliffe Hospital
Oxford, OX3 9DU, UK
email: [log in to unmask]
|