After finding that warping the T1.nii's (rather than the rT1.nii's) gave good results for the first subject I thought my problem was solved. However, I now checked the other subjects, and of the 26 subjects 5 did not return good results. I had already done a 'normal' normalization step before deciding to use DARTEL for normalization, and then these subjects did not cause problems. So I was wondering whether it is normal that the Dartel procedure sometimes doesn't work for certain subjects or that it still could have to do with the fact that I manually reoriented the scans before starting this procedure?

Kind regards, Jindra

ps. I enclosed screenshots of 2 of the 5 subjects for which it didn't work out (these are compared to the T1 template of SPM).

From: Jindra Bakker <[log in to unmask]>
Date: vrijdag 22 mei 2015 18:34
To: John Ashburner <[log in to unmask]>
Cc: "[log in to unmask]" <[log in to unmask]>
Subject: Re: [SPM] Dartel normalization problem

Dear John,

Thanks for your reply! First of all, to just be clear about what images we are talking I'll call the original image T1.nii and the manually reoriented image rT1.nii.

Secondly: I'm sorry, I wasn't clear in my last email. I used the script to see if the T1.nii and rT1.nii were aligned, and after using the script they were. I did not ask for output images with segmentation (except for default *_seg_sn.mat files of course), so couldn't check whether GM images (with which I suppose you mean output of segmentation) are aligmend with either imported GM (which I assume to be rc1* images, right?) or T1.nii's. For the last picture I send I used the rT1.nii's for normalization.

So I did a second segmentation procedure in which I did ask for c1* images. Also I did a normalization with the T1.nii's. At the images attached you see:

The template
The c1* images for one subject
The rT1.nii for that subject
The normalized rT1.nii for that subject
The T1.nii for that subject
The normalized T1.nii for that subject

The GM images are alligned with the rT1.nii's, BUT it seems that you should use the T1.nii's for normalization rather than the rT1.nii's! Do you agree? And if so: does that make sense?

Kind regards! Jindra

From: John Ashburner <[log in to unmask]>
Date: donderdag 21 mei 2015 14:01
To: Jindra Bakker <[log in to unmask]>
Cc: "[log in to unmask]" <[log in to unmask]>
Subject: Re: [SPM] Dartel normalization problem

The output from running the script would suggest that the imported GM and GM images are correct relative to each other. Perhaps you could try a Check Reg between the segmented grey matter and the original image. The script assessed whether the imported GM and GM are in alignment, but perhaps the GM and original image are not aligned. When you normalised the original image, which did you use? The reoriented one, or the un-reoriented one? Which one matches the GM image? My guess is that this one would be spatially normalised OK via Dartel.

Best regards,

-John

On 20 May 2015 at 13:57, Bakker Jindra (SP) <[log in to unmask]> wrote:

Dear John,

I had another good look at the manual and I think I followed it quite correctly. (First I manually reoriented the images.) I segmented and imported the images, I created a DARTEL template, and as you say I think it looks all right now. After that I used the Normalize to MNI space function to put the template into MNI space and warp the anatomical scans. The output of this last step is in the bottom right corner of the image I attached earlier; it is not aligned with the template, it is moved backward and rotated forward. Since I thought it might have something to do with me reorienting the images before starting the Dartel procedure, I did the following (which you suggested in https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1002&L=spm&D=0&1=spm&9=A&I=-3&J=on&K=3&d=No+Match%3BMatch%3BMatches&z=4&P=434300):
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Select images (native first, imported second)
P     = spm_select(2,'.*c1.*\.nii','Select native, then imported');

% Show the two images together
spm_check_registration(P)

% Get the secret information used by spm_check_registration
global st

% Make the hidden matrix the one that spm_check_registration uses.
st.vols{2}.mat = st.vols{2}.private.mat0;

% Something to update the view
spm_orthviews('maxbb')
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
At the last step the images were in aligment with each other, so not sure whether this really is the problem…

The other two images (top right and bottom left) are indeed original MRI images, top right after manual reorientation, bottom right before manual reorientation. To me it seems as if the output of the Normalize function did double the manual reorientation (the bottom right picture is moved twice as far to the back and oriented twice as far forwards compared to top right picture). Do you see what I mean?

I really hope you could help me with this.

Kind regards! Jindra
PhD student SMARTSCAN project | School for Mental Health and Neuroscience | Division II: Psychiatry & Neuropsychology | Maastricht University

From: John Ashburner <[log in to unmask]>
Date: dinsdag 19 mei 2015 20:36
To: Jindra Bakker <[log in to unmask]>
Cc: "[log in to unmask]" <[log in to unmask]>
Subject: Re: [SPM] Dartel normalization problem

I'm not sure what we're looking at here. The average GM in the Dartel template seems to be OK. The other images seem to be the original MRI images. Try applying the warps estimated by Dartel to these, to see if they come out aligned together. Chapter 45 of the SPM manual says more on the subject.

Best regards,
-John

On 19 May 2015 at 13:40, Jindra Bakker <[log in to unmask]> wrote: