BTW I forgot to mention that after preprocessing, CONN does not care anymore about the different slice timing, except the TR which you have to specify per subject on the Setup first tab, you can specify a vector with the TR value for each subject, eg if you have 4 subjects coming from 2 centers with the first having TR 2 and second TR 1:

[2.0 2.0 1.0 1.0]

Then since you are studying data from multiple centers, make sure at Denoising step to add global correlation regression (called gCor in CONN) as a cofound, in order to normalize the signal across centers.

If you just setup those two things in CONN, normally you can then work with CONN as if all the subjects were from one single center.

If you want additional protection, you can also add a second-level covariate to specify which center each subject is from, something like that:

centers = [1 1 2 2]

for the example above, and then at 2nd-level analysis you can use this covariate to regress out between-center differences (contrast [0 1] where 0 is centers and 1 is Allsubjects), but normally if you use gCor it should not be necessary.