The simplest way would be to regress cognitive change (controlling for important factors like age and education) w/ imaging difference maps from each participant (10 yrs minus baseline). You would want to use a conservative estimation method for the regression (e.g., TFCE or other permutation approach), plus follow-up with cross-validation (k-fold, etc.).
Another thing to consider in the neuropsych data is the sample rate of change relative to normative rate of change. If you have control participant data at both time points, then you would want to construct some sort of normative-corrected change score (e.g., reliable change indices, standardized regression-based change scores, etc.).
Warm regards,
Jeff Browndyke
Duke University Medical Center
> On Jun 22, 2021, at 12:20 PM, Marine <[log in to unmask]> wrote:
>
> Dear SPM list readers,
>
> I would like to correlate changes in brain activations with changes in memory performance (on neuropsychological tests) across two time points (1 year versus 10 years post-surgery). I have therefore, for each patient, two scans (1yr and 10yrs) and performance scores for those time points.
>
> I was wondering what would be the best way to correlate these changes? I would like to see which increases and decreases in brain activations are associated with memory changes over time.
>
> My question is, should I use a paired t-test; or should I create a new image via ImCalc for each participant, reflecting the difference between scanning sessions and then do a one-sample t-test?
>
> Many thanks in advance for your help!
>
> Marine
|