Dear John and colleagues,
Hopefully you may be able to help.
I'm investigating atrophy patterns after brain trauma, and have two brief questions about calculating annualised brain atrophy with SPM. Apologies if they are too simple! It would be great to hear your thoughts if possible.
1) Can I clarify what the JD rate values generated by SPM12's pairwise longitudinal reg tool mean? I think I can use them to give an annualised % atrophy rate but may be confused.
My understanding is that the pipeline calculates two jacobian determinants:
JD1 = Scan1 to midpoint average
JD2 = Scan2 to midpoint average
Where <0 is contraction, > 0 is growth
And the JD rate = (JD1 - JD2) / interscan-interval (eg. in years)
Is it true to state that the JD rate X 100 therefore shows the annualised % volume change at any one voxel?
Therefore if I take the mean wholebrain whitematter JD, this will show me the average % per year volume change in WM?
2) I wanted also to clarify the alternative segmentation / tissue volumes approach of answering the % volume change per year question.
For comparison with the JD approach, I am segmenting timepoints 1 & 2 and then using the tissue volumes utility to get volume values for each tissue class. I subtract the final volumes (GM/WM/CSF) from the initial and divide this by the initial scan volume. This is divided by interscan interval (years) then *100 to get annualised % volume changes.
I see that by default the tissue volumes tool uses a mask in MNI space (ie same as the TPMs for segmentation) to ensure only voxels within the brain are counted. Is it generally best to use this? Is it a problem that the input (_seg8.mat -generating-) scans are in native space?
Many thanks for your advice,
Regards,
Neil
Clinical Research Fellow
Computational, Cognitive and Clinical Neuroimaging Lab
Imperial College London
|