Dear Freesurfer developers,
I have been trying to segment (with Freesurfer v6.0.0, on Ubuntu 14.04) the hippocampal subfields of two 7T MR images in NIFTI format: an isotropic (0.8*0.8*0.8 mm3) whole-brain T1-weighted source volume with an anisotropic (0.3*1.2*0.3 mm3) T2-weighted volume covering the brain partially and centred on a region comprising both hippocampi.
I have followed the instructions in https://surfer.nmr.mgh. harvard.edu/fswiki/HighFieldRecon to correct the bias in the two 7T volumes ('T1_biasCorrected.png', 'T2_biasCorrected.png'). I have run 'recon-all -hires [...] -expert expert.txt' ('expert.txt') on my bias-corrected T1 volume, which produced an accurate segmentation ('aseg.png', 'recon-all.log'). Following this, I have run 'recon-all -hippocampal-subfields-T2' on the bias-corrected T2 volume ('hippocampal-subfields-T2.log'). Looking at the resulting segmentation, I am unsure the molecular layer has been properly delineated in the body of the hippocampus ('right_hippocampus_body.png', 'right_hippocampus_body_seg.png'). In addition, it appears the convolutions in the hippocampal head have not been correctly segmented ('right_hippocampus_head.png', 'right_hippocampus_head_seg.png').
I am not certain this is the expected behaviour when running the hippocampal subfields module on high-res data. Is it possible the recon-all output is not accurate enough (should I adjust the mres_inflate value prior to segmenting the brain)? Could the high anisotropy of the T2 volume hinder the segmentation process? Any thoughts on this? Screenshots/attachments are available on request, as my previous message was rejected for being too big.
Thanks for your help, Alexis