We previously looked at the MNI drift, and found that it is almost entirely
an affine, even for the nonlinear template. We have a suspicion that this
means that it was originally caused by the early MNI linear template, and
that the issue was known by the time MNI nonlinear templates were released,
as it implies that the nonlinear template was dedrifted, but then an affine
was deliberately added in order to match the previous MNI linear template.
However, we have not investigated whether this was the case.
Since all of our anatomical size and shape measurements are computed using
a rigid acpc space (after gradient distortion correction), this drift in
MNI space is largely unimportant to many steps of analysis of things like
fMRI. Be aware that even with proper dedrifting, any non-rigid volume
registration will hide differences in brain size (and shape if nonlinear)
across subjects, by design, so non-rigid volume-registered data won't ever
tell you the whole story of the anatomical features. If you want to make
some anatomical measurement (distance, area, volume), then you should
generally measure it in each individual's own rigid/acpc space, not in any
kind of template space.
Now, on to the explanation: the registration warpfields from each subject's
rigid acpc ("T1w") space to MNI nonlinear space provide the information
needed to compute a dedrifted template that matches the population
represented by the subjects. Warpfields are displacement vector fields
from the output coordinate system to the input coordinate system (because
of how volume resampling works - the output needs to be a regular grid, so
you actually need to transform output coordinates backwards to the original
location). So, we can take the vector fields for each subjects that
specify where the MNI-aligned features originally were in acpc space, and
average those locations for each voxel. This is simply averaging the
warpfields that are used to resample volume data from T1w/acpc space to MNI
space. Once you have this average, you can invert it (like any other
warpfield), and then concatenate it after the acpc to MNI warp for each
subject, giving you a new dedrifted (for the chosen population) template
space.
Note, however, that this requires the acpc alignment to have worked well
(in the HCP pipelines, it is really a rigid registration to an MNI
template, not specifically using the ac and pc) - if there are any
non-matching rotations among the rigidly aligned volumes, they can result
in causing a shrinkage from the true average brain volume. By carefully
finding the volume of brain masks in acpc space and after resampling to
this new dedrifted space, you can check for this effect (it will never be
an expansion). However, the effect may not be isotropic, so actually
fixing the effect properly requires more thought. However, if the acpc
alignments are generally close to being the correct orientation, this
effect should be small.
Note, in particular, that coordinates in this new space will not match MNI
coordinates, probably not even at (0, 0, 0), you would need to put any
coordinates through that same average warpfield in order to go between MNI
and this new space. You could also tweak the new space by adding a rigid
affine to the end, so you could line up (0, 0, 0) with MNI space, or
similar.
Tim
On Mon, Jul 2, 2018 at 10:21 AM, Glasser, Matthew <glass...@wustl.edu>
wrote:
> Tim Coalson can explain how to dedrift MNI space (which as Mike says is no
> longer true MNI space).
>
> Peace,
>
> Matt.
>
> On 7/2/18, 10:10 AM, "hcp-users-boun...@humanconnectome.org on behalf of
> Harms, Michael" <hcp-users-boun...@humanconnectome.org on behalf of
> mha...@wustl.edu> wrote:
>
> >
> >Hi,
> >Re (1):
> >If you want to work with streamlines in MNI space then you have to accept
> >the "drift" (volume expansion) that is part of MNI space itself.
> >
> >Re (2):
> >Maps may be resampled to surfaces in MNI space. But all the FreeSurfer
> >structural quantities that we provide in the database/spreadsheet in
> >ConnectomeDB are measured in the subject's real (native) space.
> >
> >Cheers,
> >-MH
> >
> >--
> >Michael Harms, Ph.D.
> >
> >-----------------------------------------------------------
> >
> >Associate Professor of Psychiatry
> >
> >Washington University School of Medicine
> >
> >Department of Psychiatry, Box 8134
> >
> >660 South Euclid Ave. Tel: 314-747-6173
> >
> >St. Louis, MO 63110 Email: mha...@wustl.edu
> >On 7/2/18, 7:14 AM, "hcp-users-boun...@humanconnectome.org on behalf of
> >Claude Bajada" <hcp-users-boun...@humanconnectome.org on behalf of
> >c.baj...@fz-juelich.de> wrote:
> >
> >Dear experts,
> >
> >I have a few questions:
> >
> >1) I am currently working on a project using tractography and the HCP
> >data. I have used the HCP provided FNIRT based MNI registration in order
> >to register individual streamlines from "native" or diffusion space (as
> >per post HCP minimal processing pipeline) to MNI space. However, in
> >Glasser 2016 supplementary text, it states that the MNI space is drifted
> >relative to the mean brain size of individual brains (greatest along the
> >x and z axes). I've been told that it is possible to "dedrift" these
> >data and I wonder if anyone has experience in doing this and is able to
> >point me in the right direction.
> >
> >2) from what I understand, structural qualities such as cortical
> >thickness etc are measured in the individual's real space. However then,
> >what I do not understand is the file structure of the HCP data. The
> >structural measures such as cortical thickness, curvature and myelin
> >maps etc live in the MNINonLinear/fsaverage_LR32k rather than in the
> >'native space' directory. Can anyone shed light on this?
> >
> >Regards,
> >
> >Claude
> >
> >
> >
> >-----------------------------------------------------------
> ---------------
> >----------------------
> >-----------------------------------------------------------
> ---------------
> >----------------------
> >Forschungszentrum Juelich GmbH
> >52425 Juelich
> >Sitz der Gesellschaft: Juelich
> >Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498
> >Vorsitzender des Aufsichtsrats: MinDir Dr. Karl Eugen Huthmacher
> >Geschaeftsfuehrung: Prof. Dr.-Ing. Wolfgang Marquardt (Vorsitzender),
> >Karsten Beneke (stellv. Vorsitzender), Prof. Dr.-Ing. Harald Bolt,
> >Prof. Dr. Sebastian M. Schmidt
> >-----------------------------------------------------------
> ---------------
> >----------------------
> >-----------------------------------------------------------
> ---------------
> >----------------------
> >
> >
> >_______________________________________________
> >HCP-Users mailing list
> >HCP-Users@humanconnectome.org
> >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.
> >
> >_______________________________________________
> >HCP-Users mailing list
> >HCP-Users@humanconnectome.org
> >http://lists.humanconnectome.org/mailman/listinfo/hcp-users
>
>
> _______________________________________________
> HCP-Users mailing list
> HCP-Users@humanconnectome.org
> http://lists.humanconnectome.org/mailman/listinfo/hcp-users
>
_______________________________________________
HCP-Users mailing list
HCP-Users@humanconnectome.org
http://lists.humanconnectome.org/mailman/listinfo/hcp-users
