First of all, ensure that you have a relatively up to date version of SPM8.  There was a bug in the original version of ``Normalise to MNI'', which was fixed in a more recent update.

I had similar problem when applying DARTEL spatially normalization of functional data to MNI space as described in SPM8 manual, Chapter 41.2. Prior to the coregistration, structural and functional images were reoriented via Display in order to have better starting estimates, and functional images were realigned.

I've performed DARTEL  routine several times:

1.    After coregistration of structural scan with mean*.nii resulted from realignment.

2.    After structural scans were scull-stripped and functional realigned scans “moved” to the scull-stripped structural scans (Coregistration: Reference image - structural scull-stripped scan, Source image - mean*.nii, Other images - all realigned functional scans).

These both seem to be about right.

Depending how the skull stripping is done (perhaps if you used software that doesn't know about NIfTI format), this may result in the skull stripped images not being in alignment with the originals.  Use Check Reg to make sure this isn't the case.

Importantly, (1) coregistration was successful for all options described above, (2) the structural scans as well as c1*.nii and c2*.nii were successfully normalized to MNI space. However, activation maps did not fit well MNI space - they are smaller.

Did the spatially normalised fMRI fit MNI space well?  If there are regions that don't fit MNI space in the warped fMRI data, which appear to be OK in the structural data, then it is perhaps worth checking that part of the brain in the coregistered fMRI and structural scans (Check Reg again).

I'd also suggest taking a look at the mask.img, comparing this with MNI space (Check Reg).  Perhaps the ``GM masking'' has chopped out part of the data.

Finally, I’ve normalized functional data using  EPI_template applied by SPM. Activation maps obtained using the same model specification and estimation routine with these images, fit the MNI space very well. Therefore, masking threshold in spm_defaults.m is not a reason for these discrepancies.

That seems to be a reasonable conclusion, but it may still be worth checking the mask.img.  All it takes is for one subject to be poorly aligned for some reason, and the mask will be changed considerably.
 

Maybe a clue to this problem can be found in the following notion:

“… when applying a GLM to the imported data, the information in the second matrix will not be written to the NIfTI headers of the beta and contrast images.  This means that DARTEL won't apply the right warps.” -  from  https://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind0904&L=SPM&P=R7426

This one seems to be an unlikely explanation here.
 

Perhaps John can shed some light on this issue.

I hope this has helped.

Best regards,
-John

 

On Wed, Feb 6, 2013 at 10:02 PM, Neggers, S.F.W. <[log in to unmask]> wrote:

Dear Bianca,

How did you set the masking threshold in spm_defaults.m in both versions? That can make quite a difference. Your images hint at that making the difference rather than registration.

To check whether registration is the issue, these plots are not very helpful. Do a check-reg with 1 preprocessed functional (on which you run your GLM that is) and your preprocessed T1. They should match. That is always a very good check to do for every subject, by the way, before doing GLM. If these 2 scans dont match, registration at some point during your pipeline went wrong, otherwise something else like your masking threshold might be the culprit.

Finally, without posting your full preprocessing pipeline, information about the acquisition, and which part of SPM5 vs SPM8 you think is 'better' it is pretty hard to give you good advice. As far as I know mainly normalization (through unified segmentation) improved somewhat from SPM5 to SPM8, but realignment, coreg, stayed relatively constant from my experience. But perhaps John can comment on that in more detail.

Good luck,

Bas


--------------------------------------------------
Dr. S.F.W. Neggers
Associate Professor, Brain Division
Rudolf Magnus Institute for Neuroscience
Utrecht University Medical Center

Visiting :  Heidelberglaan 100, 3584 CX Utrecht
                Room B.01.1.03
Mail       :  Huispost B01.206, P.O. Box 85500
                 3508 GA Utrecht, the Netherlands
Tel        : +31 (0)88 7559609
Fax        : +31 (0)88 7555443
E-mail   : [log in to unmask]
Web      : http://www.nstarlab.nl (PI)
              : http://www.brainsciencetools.com (CEO)
--------------------------------------------------
________________________________________
From: SPM (Statistical Parametric Mapping) [[log in to unmask]] on behalf of Bianca van den Bulk [[log in to unmask]]
Sent: Wednesday, February 06, 2013 4:10 PM
To: [log in to unmask]
Subject: [SPM] registration problem SPM8

Dear SPM'ers,

I hope someone can give me some advice about the following problem we ran into:

Last month we decided to rerun all our analyses in SPM8 instead of SPM5. On reason for this is that the registration process is supposed be better compared to the registration in SPM5
After doing all the analyses we noticed a difference in the whole brain activation patterns, especially in the visual cortex.
With SPM5 the visual cortex activation neatly fits with the outside of the visual cortex. However, in SPM8 this is not the case. I attached two images (one from spm5 and one from spm8) in which the difference is visible. Both pictures are made with the same group of participants and with the same contrast and therhold.
In my oppinion the brain seems to be shrinked in the spm8 image (both on the posterior and anterior side of the brain)...

To check whether the registration is the problem or not, we also ran the SPM8 analyses without registration to T1 scan. Then the rendered images correspond to the images created with SPM5. To us this suggests that something goes wrong with the registration from functional data to T1 scan to template. However, we do not know what goed wrong.

Does someone have an idea about this problem?

Thanks!

Bianca

------------------------------------------------------------------------------

De informatie opgenomen in dit bericht kan vertrouwelijk zijn en is
uitsluitend bestemd voor de geadresseerde. Indien u dit bericht onterecht
ontvangt, wordt u verzocht de inhoud niet te gebruiken en de afzender direct
te informeren door het bericht te retourneren. Het Universitair Medisch
Centrum Utrecht is een publiekrechtelijke rechtspersoon in de zin van de W.H.W.
(Wet Hoger Onderwijs en Wetenschappelijk Onderzoek) en staat geregistreerd bij
de Kamer van Koophandel voor Midden-Nederland onder nr. 30244197.

Denk s.v.p aan het milieu voor u deze e-mail afdrukt.

------------------------------------------------------------------------------

This message may contain confidential information and is intended exclusively
for the addressee. If you receive this message unintentionally, please do not
use the contents but notify the sender immediately by return e-mail. University
Medical Center Utrecht is a legal person by public law and is registered at
the Chamber of Commerce for Midden-Nederland under no. 30244197.

Please consider the environment before printing this e-mail.

--

Ella Gabitov

Ph.D student

The E.J. Safra Brain Research Center for Learning Disabilities

The Laboratory of Human Brain and Learning

University of Haifa

Israel

Phone: +972-52-6722700

E-mail: [log in to unmask]