Dear SPM'ers,
a colleague gave me the solution. I finally used Imcalc to get the difference images.
Thanks,
Nicole
----- Original Message -----
From: "Nicole Hudl" <[log in to unmask]>
To: [log in to unmask]
Sent: Thursday, October 27, 2016 6:18:55 PM
Subject: Longitudinal VBM, multiple regression
Dear SPMers,
I am writing to you, because I have a problem with the analysis of my structural data. I need some words to explain my study before I can ask my question. So please be patient.
I want to investigate the effect of a cognitive training on structural plasticity. In detail, I am interested, whether there are grey matter density changes from pre-training to post-training scan that can directly be related to the success in the training (operationalized by training level). Since I can not compute pre-vs-post contrast images using structural images at individual level (or I haven't found it yet), I tried two approaches:
First, I modelled a 2-sample-t-test with pre-scans as first group and post-scans as second group. As the "groups" (two time points of same individual) are not independent, I set this parameter to dependent. Then, I created two covariates, the first called "training level session 1" and a second one named "training level session 12" which is the start and the end of the training period.
The first covariate was filled with the last level reached at the end of the first session. Since "Group2" represents the post-training scan, I filled in zeros. For the second covariate, I first filled in zeros for "Group1" and then added the last level reached at the end of session 12.
I corrected for age and TIV. In the contrast manager, I tested for example:
Group 1 Group 2 TIV AGE training level session 1 training level session 12
1 -1 0 0 -1 1
looking for grey matter density decreases related to training level increases (since all participants significantly increased their level from session to session).
Since my supervisor was not sure whether it is appropriate to use the 2-sample-t-test for a within-group design (even though I set independence to no), we decided to design a paired t-test, using the pre-training and the post-training scan as a pair for each participant. Therefore, I created one covariate "training level" and added the corresponding numbers.
I corrected for age and TIV. In the contrast manager, I tested for example.
Group 1 Group 2 TIV AGE training level
1 -1 0 0 1
looking for grey matter density decreases that are positively related to training level.
I got totally different results and I am not sure whether one of the method is testing what I want and if so, which method is the one to use.
I haven't found a way to do multiple regression in a longitudinal VBM and I hope to find some help here.
Thanks in advance,
Nicole
Nicole Hudl, M.Sc. Psychology
PhD student
Fellow of the MaxNetAging Research School
Max Planck Institute for Human Cognitive and Brain Sciences
Department of Neurology; Day Clinic for Cognitive Neurology
|