I would like to provide some context on this issue, as I have worked
on it in a few papers.
The first documented discussions in the neuroimaging community on the
inadequacy of T1-weighted images alone to stake out gray matter stem
from two sources.
The first are discussions in the Freesurfer community. In Freeserfer,
you have to manually edit the segmentation prior to the generation of
the cortical thickness estimates. If you have ever tried to do this,
you will be familiar with the (meningeal) dura problem: the meningeal
dura is clearly separated from cortex, but fuses with it at some point
and it becomes visually indistinguishable. The Freesurfer group first
proposed to handle this by using T2* weighted images, in which dura
gives a different contrast from gray matter. A rationale for this is
that T2* images can be obtained practically for free if one records
the echos in the MPRAGE sequence (multiecho MPRAGE).
A second source is a paper by Helms et al. showing that T1-weighted
images do a bad job at individuating the cortex by comparing them with
the signal from T2-weighted images. Hence, one immediate workout is to
use both for segmentation. This comes at a cost, however, as the
T2-weighted images must be acquired separately. One thing that is
clear from this paper is that not only dura, but also subarachnoid
vessels are confounded with cortex in MPRAGE images.
In the Freesurfer community (I mean, what they say on their mailing
list) there was at some point a switch from using T2* to combining the
segmentation with the data from a FLAIR image (T2-weighted). As far as
I know, they did not document the advantages of FLAIR as they had done
with the T2* technique.
If you compare the effectiveness of the [multiecho MPRAGE + T2*] with
the [MPRAGE + FLAIR] strategies, it turns out that the latter is much
more effective (Viviani et al. 2017a). This is not suprising, as the
FLAIR brings additional signal collected in the time it took to
acquire it, in contrast to the T2*. T2* is very noisy, and also
suffers from large susceptibility effects. It also turns out that the
combined segmentations correct for the vessels confound, as first
suggested by Helms et al.
This is not all, however. The intensity of the cortex is not uniform
across the mantle due to variations in the intensity of cortical
myelin (Eickhoff et al. 2005). Areas that are myelin rich are mainly
primary sensory and motor areas, and visual areas in the occipital and
clalcarine cortex. MPRAGE-based segmentations tend to underrepresent
this cortex, especially in the visual areas. The [MPRAGE+FLAIR]
combination does a much better job here too. In fact, it may be shown
that this combination contains the signal to discriminate the
myelinated layers of cortex (Viviani et al. 2017b).
Best wishes,
Roberto Viviani
References
Eickhoff, S., Walters, N.B., Schleicher, A., Kril, J., Egan, G.F.,
Zilles, K., Watson, J.D.G., Amunts, K., 2005. High-resolution MRI
reflects myeloarchitecture and cytoarchitecture of human cerebral
cortex. Hum. Brain Mapp. 24, 206–215.
Helms, G., Kallenberg, K., and Dechent, P. (2006). Contrast-driven approach to
intracranial segmentation using a combination of T2- and T1-weighted 3D
MRI data sets. J. Magn. Reson. Imaging 24, 790–795. doi: 10.1002/jmri.20692
van der Kouwe, A. J. W., Benner, T., Salat, D. H., and Fischl, B. (2008).
Brain morphometry with multiecho MPRAGE. NeuroImage 40, 559–569.
doi: 10.1016/j.neuroimage.2007.12.025
Viviani, R., Pracht, E.D., Brenner, D., Beschoner, P., Stingl, J.C.,
Stöcker, T., 2017a. Multimodal MEMPRAGE, FLAIR and R2* segmentation to
resolve dura and vesslels from cortical gray matter. Frontiers in
Neuroscience doi: 10.3389/fnins.2017.00258
Viviani, R., Stöcker, T., and Stingl, J. C. (2017b). Multimodal FLAIR/MPRAGE
segmentation of cerebral cortex and cortical myelin. NeuroImage 152, 130–141.
doi: 10.1016/j.neuroimage.2017.02.054
Zitat von "Stewart, Peter" <[log in to unmask]>:
> Thanks for this article, I?ve been wondering about this for quite
> some time and haven?t found any systematic study (there are a few
> dissertations floating around but nothing more formal until this
> article).
>
> Peter
>
> From: SPM (Statistical Parametric Mapping) <[log in to unmask]> On
> Behalf Of Kotikalapudi Raviteja
> Sent: Monday, January 21, 2019 10:22 AM
> To: [log in to unmask]
> Subject: [External] Re: [SPM] multimodality image segmentation
>
> SPM12-based multimodal segmentation does show improvement over
> conventional T1w-only method. You can see below the link for the
> systematic comparison of different multiple-channel methods using
> SPM12 below:
> https://www.sciencedirect.com/science/article/pii/S1053811917301209<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.sciencedirect.com%2Fscience%2Farticle%2Fpii%2FS1053811917301209&data=02%7C01%7Cpstewart1%40GEISINGER.EDU%7C9550c30180c04bad34a008d67fb52764%7C37d46c567c664402a16055c2313b910d%7C0%7C0%7C636836813358104386&sdata=msLhXEYrOcJrsN07k0joQkIjf06gAqllQvtFpP2VuKY%3D&reserved=0>
>
> In brief, issues such as overestimating dura/vessels are addressed
> through multichannel segmentation (T1+T2w) as it often gets
> segmented as GM in T1w-only approach.
>
> Best,
> Ravi
>
> On Mon, Jan 21, 2019 at 1:43 PM Wiktor Olszowy
> <[log in to unmask]<mailto:[log in to unmask]>> wrote:
> Dear Barbara,
>
> I was just wondering: which segmentation eventually worked better
> for you: "SPM12 Segment using the multiple channels (T1 and
> T2-weighted data) or CAT12 with only T1w- data"?
>
> Best wishes,
>
> Wiktor Olszowy
>
>
> IMPORTANT WARNING: The information in this message (and the
> documents attached to it, if any) is confidential and may be legally
> privileged. It is intended solely for the addressee. Access to this
> message by anyone else is unauthorized. If you are not the intended
> recipient, any disclosure, copying, distribution or any action
> taken, or omitted to be taken, in reliance on it is prohibited and
> may be unlawful. If you have received this message in error, please
> delete all electronic copies of this message (and the documents
> attached to it, if any), destroy any hard copies you may have
> created and notify me immediately by replying to this email. Thank
> you. Geisinger Health System utilizes an encryption process to
> safeguard Protected Health Information and other confidential data
> contained in external e-mail messages. If email is encrypted, the
> recipient will receive an e-mail instructing them to sign on to the
> Geisinger Health System Secure E-mail Message Center to retrieve the
> encrypted e-mail.
|