Dear Vasudev,

It is expected that you get different volume estimates with different atlases (FreeSurfer vs CobraLab), as the manual delineation protocols that were used to create them are very different (e.g., see Yushkevich et al., NeuroImage, 2015).

Regarding the differences in p-values between the two methods:

- Do you get a similar effect size / pval when looking at the whole hippocampus?

- Are the volumes of the subfields given by one method highly correlated with the volumes given by the other?

- Did you QC the output, for both methods?

- How different were the results you got? While you mention statistical significance vs not, I wouldn’t be alarmed if the p-value was 0.04 vs 0.06, whereas 0.00001 vs 0.5 would be much more worrying.

Cheers,

/Eugenio

Juan Eugenio Iglesias

ERC Senior Research Fellow

Centre for Medical Image Computing (CMIC)

University College London

http://www.jeiglesias.com

http://cmictig.cs.ucl.ac.uk/

From: <freesurfer-bounces@nmr.mgh.harvard.edu> on behalf of Dev vasu <vasudevamurthy.devulapally@gmail.com>
Reply-To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edu>
Date: Thursday, 2 August 2018 at 19:18
To: "freesurfer@nmr.mgh.harvard.edu" <freesurfer@nmr.mgh.harvard.edu>
Subject: [Freesurfer] Anterior / Posterior Partition in Hippocampus parcellation atlas

External Email - Use Caution

Dear All,

I have a problem with the Freesurfer hippocampal longitudinal segmentation results , when i have done segmentation using following algorithm ( https://github.com/CobraLab/antsRegistration-MAGeT) with no anterior/posterior partitioning ( Head/body/tail ) , on my subject pool the volumes that i have obtained for each subfield is very different from the volumes of corresponding subfields that i am getting when i have done subfield segmentation using freesurfer. I think algorithm does influence the segmentation procedure, if this is the case then i am getting different results for histology based segmentation using https://github.com/CobraLab/antsRegistration-MAGeT and freesurfer.

When i compare the subfield volumes between healthy and disease patients that i have obtained using first segmentation procedure using ANTS-MAGeT , i notice statistical significance in subfields like CA1,CA2-CA3 but when i do the same using freesurfer based algorithm i notice no statistical significance in these subfields.

Do you think the algorithm has any influence on segmentation to such a level that we get completely different results on two occasions.

Thanks

Vasudev