Dear FSL Experts,
We have undertaken a longitudinal fMRI study (block design) where participants have been scanned at four separate time points: a pre-intervention scan, and three post-intervention scans at 4, 8 and 12 months (the intervention was only applied once, shortly after the pre-intervention fMRI scan). We have performed a first level FEAT analysis for all participants (20 contrasts of interest) at each of these time points and registered to the standard MNI brain template. I am having difficulty in identifying the appropriate way to perform a higher level group analysis and would like to ask advice on the following please:
1) What is the appropriate design matrix for the group analysis for such a longitudinal study with the four fMRI scan time points? From the FSL FEAT page, I thought the example given which looked like the one to select for a longitudinal design was the “F-Tests”. I attempted such a group analysis by selecting six participants, each of whom had all four fMRI scans. Please see image attached of the design matrix and contrasts.
Does this group design matrix look correct and am I using it appropriately in the context of a longitudinal fMRI study? Although not stated on the FSL FEAT page, I simply performed contrasts between the various scan points e.g. pre-intervention scan > 12 months scan using [1 -1] in the contrasts (please see attached). My instinctive feeling is that this is probably not the appropriate way to do this (need to take into account repeated measures?) so I would be very grateful for guidance here.
2) Another issue that has arisen in the longitudinal fMRI study is that not all the participants were able to complete all their post-intervention fMRI scans, and data from one pre-intervention fMRI scan is probably not usable. For example, some participants have missed scans at the 4 month time point and others at 8 and/or 12 months. How should we deal with these “missing” fMRI datasets in the group design matrix for a longitudinal fMRI study?
3) Lastly, is it possible to add covariates into the longitudinal group design matrix for each of the fMRI scans? If yes, how should this be done?
With best wishes,
Aziz Asghar.
|