Greetings all,
I would like to use deformation-based and voxel-based morphometry together to identify macroscopic (e.g., shape, relative position) and mesoscopic (e.g, local tissue volume / density) predictors of treatment outcomes. To run the DBM analysis, I first used the VBM8 toolbox to generate a affine-transformed, skull-stripped partial-volume labeled image for each subject. Next, I re-ran the toolbox to generate nonlinear deformation fields for each subject, using the PVE-image as input. After processing, I included the deformation fields for each subject in a 2nd level model looking at treatment group x treatment outcome interactions, also including age, sex and education as regressors of no interest. A couple of questions for the stats experts:
1) Considering that the deformation fields were generated from affine-transformed anatomical images, where I see significant differences between groups, can I make inferences about local structural shape?
2) Do the same issues of non-stationarity apply for analysis of deformation fields as they do for VBM?
3) Any advice for improving this analysis?
Thanks in advance for any comments/suggestions.
All the best,
Patrick
|