Yes, the skull is only used for the initial, low resolution, iterations. It is useful there as it sometimes "stabilises" things and prevents fnirt from finding some silly local minima while it is still far from the "global" minimum.
After the we use a brain mask that basically covers the standard brain and a couple of voxels outside the brain (to make sure we get the edge). That means that for the rest of the iterations the skull can go pretty much anywhere as dictated by the warps on the brain surface and the regularisation function. The reasons for this is i) We don't want to "waste" warp resolution by reconciling warps on the brain surface and the skull and ii) We find that the variability (across scanners and sequences) in signal is greater outside the brain than inside. Probably because physicists don't really care what is going on outside the brain when they optimise sequences/protocols. If you do want to revisit the phrenology field you can easily turn the masking off in the config file. I can also recommend the phrenology museum in Edinburgh. Very fascinating. Jesper On 16 Sep 2013, at 17:33, "Glasser, Matthew" <[email protected]> wrote: > Yes, given our extensive phenotypic characterization of the HCP subjects, > we can now say that bumps over the posterior temporal lobe are related to > better language function and over the parietal lobe to better spatial > abilitiesŠ :) > > More seriously, I think you are right about what is causing this. We > didn't do any special optimization of the FNIRT config for the HCP data, > and thus are using the default T1_2_MNI152_2mm.cnf file. This does > include the MNI space brain mask, which I believe sets the cost function > weighting. Mark and Jesper would know more about how FNIRT works > internally than I do. > > We're mainly relying on FNIRT to align subcortical structures, however, as > cortical structures are better aligned with surface-based methods. You > can see in the attached images (coefficient of variation of the T1w image) > that FNIRT does very well in the subcortical grey structures. > > Best, > > Matt. > > On 9/16/13 10:04 AM, "Robert Cox" <[email protected]> wrote: > >> I started looking at some of the T1w volumes from the 2013Q2 release. >> In 2 cases that I noticed, the warped volumes in these subjects' >> MNINonLinear/ >> directory show marked bumpiness in the skulls. At first I thought, >> "phrenologist's >> delight", but then I checked the original volumes -- alas, no >> particular bumpiness >> there. So presumably this bumpiness was induced by the nonlinear >> registration. >> The extracted brain-only volumes (T1w_restore_brain.nii.gz) don't look >> as lumpy. >> Is the nonlinear warping goodness-of-fit measure confined to the brain >> volume, >> and this artifact is induced by edge effects? Or what's going on? >> >> Cases of note are 581349 and 133928. There may be more such lumpy >> skulls, since >> this was just a quick perusal. >> >> -- bob cox >> _______________________________________________ >> HCP-Users mailing list >> [email protected] >> http://lists.humanconnectome.org/mailman/listinfo/hcp-users > > > ________________________________ > The materials in this message are private and may contain Protected > Healthcare Information or other information of a sensitive nature. If you are > not the intended recipient, be advised that any unauthorized use, disclosure, > copying or the taking of any action in reliance on the contents of this > information is strictly prohibited. If you have received this email in error, > please immediately notify the sender via telephone or return mail. > <Q1-2_Related120_T1w_COV_z=0.png> _______________________________________________ HCP-Users mailing list [email protected] http://lists.humanconnectome.org/mailman/listinfo/hcp-users
