Dear Franz,
I still need to see the corresponding T1-image to make any statement, whether this location should be GM or WM. In your PDF 3 different segmentations approaches with 3 different normalization methods (DARTEL/SPM8/New Segment) are mixed and the localizations cannot be directly compared.
If you intend to save the normalized T1 image with VBM8, please enable the option VBM8|Writing options||Bias Corrected|Normalized. This will save the T1 image even if you use the default DARTEL normalization. An alternative approach to check the quality of the segmentations is to save the segmentations and the T1 image in native space to preserve the original spatial resolution.
Regards,
Christian
____________________________________________________________________________
Christian Gaser, Ph.D.
Assistant Professor of Computational Neuroscience
Department of Psychiatry
Friedrich-Schiller-University of Jena
Jahnstrasse 3, D-07743 Jena, Germany
Tel: ++49-3641-934752 Fax: ++49-3641-934755
e-mail: [log in to unmask]
http://dbm.neuro.uni-jena.de
On Thu, 18 Nov 2010 23:24:54 +0100, Franz Riederer <[log in to unmask]> wrote:
>Dear Christian,
>
>insisting is good!
>
>Thank you for the DARTEL-lesson. To my experience the DARTEL-VBM workflow
>does not create a DARTEL-normalised T1, I appologize I did not attempt to
>create one.
>
>Please let me explain my question with the following image:
>
>Top row left:
>modulated grey, VBM8 defaults with the following exceptions: extremely light
>bias regularisation (0.00001), SPM-default modulation.
>
>Top row right:
>Corresponding WM
>
>Lower row left:
>SPM8 New segment, extremely light bias regularisation (0.00001), otherwise
>defaults
>
>
>- right: SPM8 segment, extremely light bias regularisation (0.00001),
>thorough clean up, otherwise defaults.
>
>
>From this comparison, I got the impression that some GM might be missing in
>the VBM8 segmentation (and to a lesser extent also using New segment).
>(Please see the WM segment here). Inspecting the native volume, I saw that
>the dura was rather bright here, with similar intensity as WM.
>I was considering a partial volume effect, but the white matter seems to
>reach the surface of the brain.
>
>What do you think?
>
>Thanks a lot
>
>
>Franz
>
>
>PS: Thanks also to Julie and John for their valuable comments.
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>2010/11/17 Christian Gaser <[log in to unmask]>
>
>> Dear Franz,
>>
>> sorry for insisting, but the normalized T1-image in your pdf-file is still
>> not in the same space as the segmented images (please check some single gyri
>> and sulci in your images). VBM8 uses DARTEL normalization as default, which
>> differs largely from the default SPM8 normalization.
>>
>> Regards,
>>
>> Christian
>>
>> On Wed, 17 Nov 2010 10:13:05 +0100, Franz Riederer <
>> [log in to unmask]> wrote:
>>
>> >Dear Christian,
>> >
>> >In the second figure I attached, the normalised T1 is included.
>> >
>> >I observed that not all brain surface appears to be surrounded by grey
>> >matter. I was thinking of a partial volume effect first , but when you
>> look
>> >at the corresponding white matter, you see that it seems to "fill" the
>> dark
>> >spots in the grey matter segment (first figure).
>> >
>> >Thanks for any remedy,
>> >
>> >Franz
>> >
>> >
>> >
>> >
>> >2010/11/17 Christian Gaser <[log in to unmask]>
>> >
>> >> Dear Franz,
>> >>
>> >> my first guess is that the GM/WM images are normalized, but the T1 image
>> is
>> >> not. You have to use the normalized T1 image to correctly compare it
>> with
>> >> its segmented counterparts.
>> >>
>> >> Regards,
>> >>
>> >> Christian
>> >>
>> >>
>> >>
>> ____________________________________________________________________________
>> >>
>> >> Christian Gaser, Ph.D.
>> >> Assistant Professor of Computational Neuroscience
>> >> Department of Psychiatry
>> >> Friedrich-Schiller-University of Jena
>> >> Jahnstrasse 3, D-07743 Jena, Germany
>> >> Tel: ++49-3641-934752 Fax: ++49-3641-934755
>> >> e-mail: [log in to unmask]
>> >> http://dbm.neuro.uni-jena.de
>> >> On Tue, 16 Nov 2010 14:44:25 +0100, Franz Riederer <
>> >> [log in to unmask]> wrote:
>> >>
>> >> >Dear all,
>> >> >
>> >> >I see there is a lot of effort to support people working with
>> structural
>> >> MRI
>> >> >analyses, and I am grateful for that. Although I got some inspiration
>> from
>> >> >the archives of the mailing list, I could not find an answer for the
>> >> >following problem:
>> >> >In some rare instances grey matter seems classified as white matter
>> >> (please
>> >> >see attachment, blue arrow). In these instances, the dura over the
>> cortex
>> >> >appears rather bright (lower image). Changing the bias regularisation
>> to
>> >> >"extremely light" instead of the default "very light" seemed to have a
>> >> minor
>> >> >positive effect.
>> >> >
>> >> >To my surprise, the Unified Segmentation within SPM8 seemed to be less
>> >> >sensitive to this artifact.
>> >> >
>> >> >I also see that this might be a problem of the sequence we used in this
>> >> >subject.
>> >> >
>> >> >I wonder whether there is a way to optimize post processing.
>> >> >
>> >> >Many thanks in advance,
>> >> >
>> >> >
>> >> >Franz
>> >> >
>> >>
>> >>
>> >>
>> >
>>
>>
>>
>
|