Hi,
I am looking at brain development in a sample of 500 adolescents, aged 17 to 26. Typically, at this age, the cortex thins out a bit (see Tamnes, JNneuroscience 2017 for example), and that is what we observe using Freesurfer (longitdudinal model). I wanted to confirm this using SPM longitudinal registration, did the following procedures:
SPM12(7771) Longitudinal registration
entered all the time 1 images
entered all the time 2 images
entered the time in years (ranging from 0.1 to 3.5 years)
Saved average image, DV and Jacobians
Segmentation and Dartel of the midpoint images
Normalized DV images to MNI, 8 mm smoothing
second level regression model on sw_dv_time_a_time_b.nii, including age, sex, ICV
Modeling the first regressor (constant) gives me a nice map of expansions and contractions, but the frontal cortices expand (positive dv), while the parietal cortex contracts (negative dv), and sensory cortices stay the same
I also tried the approach using Jacobins and jacobians*c1, or dv*c1 before normalizing, but with similar effects.
To verify the apparent volume increase, I also segmented the time1 and the time2 images independently, and calculated tissue volumes. Also here, I see an average reduction in gray matter volume.
Any suggestions to why this tensor-based morphometry gives different results than freesurfer cortical thikness? Could it be the "opposite" results are due to development of the gray-white matter border?
|