External Email - Use Caution
Hello Douglas,
Thank you for your advice.
As you have suggested, I have now used orig.mgz from the recon-all longitudinal processing stream as an input to Samseg longitudinal processing stream. For comparison, I have the following volumes
1) Left-Hippocampus volume from recon-all longitudinal processing stream in aseg.stats (3321.5, mm^3) 2) Left-Hippocampus volume from samseg longitudinal processing stream in samseg.stats (3660.103030, mm^3) 3) Left-Hippocampus volume from longitudinal Hippocampal subfields processing stream in lh.hippoSfVolumes-T1.long.v21.txt (Whole_hippocampus 3012.029850, mm^3; Whole_hippocampal_body 999.162202, mm^3; Whole_hippocampal_head 1581.449339, mm^3; Hippocampal_tail 431.418310, mm^3)
When I visually check, the Left-Hippocampal segmentation appears to be better from recon-all longitudinal processing stream. There appears to be a slight overestimation with samseg longitudinal processing stream.
Question 1) Additionally, the longitudinal Hippocampal subfields processing stream gives another set of volumes. Which set of Left-Hippocampal subfields volumes would correspond to Left-Hippocampal volume from recon-all longitudinal processing stream?
Question 2) If i want to analyse the Hippocampal subfields volumes, would it make sense to correct for ICV from the samseg longitudinal processing stream or the hippocampal volume from recon-all longitudinal processing stream?
BW Jay
On Mon, Feb 22, 2021 at 4:14 PM Douglas N. Greve dgreve@mgh.harvard.edu wrote:
On 2/18/2021 7:06 AM, Jayachandra Raghava wrote:
External Email - Use CautionHello Freesurfer Experts,
I am using Freesurfer version 7.1.1 to perform longitudinal Samseg processing (using run_samseg_long) on my T1 longitudinal data.
- First, I run the mri_robust_template to generate the mean.mgz using
T1.mgz from the cross sectional recon-all pipeline. Would it be better to use T1.mgz from the longitudinally processed data?
You mean the T1.mgz from the recon-all output? Don't use that. Use the orig.mgz.from the long stream (you'd only need to run it long enough to produce the orig.mgz)
- There are few subjects who do not have followup data. How do I process
these subjects to avoid the bias introduced by the longitudinal Samseg pipeline?
You can run the recon-all long stream on that one time point; it will apply a registration designed to not introduce bias, then use the orig.mgz
- For the same subject, I see that the subcortical volumes from
longitudinal recon-all pipeline from aseg.stats (e.g Left-Hippocampus volume is 3321.5mm3) are different to the subcortical volumes from longitudinal samseg pipeline from samseg.stats (e.g Left-Hippocampus volume is 3899.628686mm3). Which of these measures would be recommended?
In general I would say samseg, but that is a pretty big difference. Can you take a look and see if one looks better than the other?
Thank you very much for your support!
Best wishes Jay
Freesurfer mailing listFreesurfer@nmr.mgh.harvard.eduhttps://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://secure-web.cisco.com/1fZIfuPY-paBVpw9ZnsGEuquODX-06EZY9uM_PB7S14dDUZ...