Hello,
I have added my 2 cents several times on this subject but it does not
look like this was ever really conclusively answered. As back in May
2009, I still think the answer is "not much" :) but this will always
somewhat depend on the dataset under question, so if you want to look
for yourself, you can investigate the effects of using the different
regularizations in your sample with the following snippet of code (watch
out for linebreaks when copying this to your Matlab window):
%=== start ===
imgs = spm_select(Inf, 'image', 'Select images to test');
tpm = spm_load_priors8(spm_select(1, 'image', 'Select TPM'));
nimgs = size(imgs,1);
opts = char('mni', 'eastern', 'rigid', 'subj', 'none');
nopts = size(opts,1);
sep = 3;
fwhm = 8;
store = zeros(nimgs, nopts, 4);
for i = 1:nimgs
for ii = 1:nopts
M = spm_maff8(imgs(i,:),sep,fwhm,tpm,[],deblank(opts(ii,:)));
store(i,ii,1) = det(M); % Matrix determinant
M = spm_imatrix(M);
store(i,ii,2) = M(7); % scaling in X
store(i,ii,3) = M(8); % scaling in y
store(i,ii,4) = M(9); % scaling in z
end;
end;
clc;
disp(' The characteristics of the affine transformation matrix
are...');
for i = 1:nimgs
for ii = 1:nopts
disp([' det = ' sprintf('%01.4f', store(i,ii,1)) ', scal(x) = '
sprintf('%01.4f', store(i,ii,2)) ', scal(y) = ' sprintf('%01.4f',
store(i,ii,3)) ', scal(z) = ' sprintf('%01.4f', store(i,ii,4)) ]);
disp([' for option ' deblank(opts(ii,:)) ' for image ' num2str(i) ]);
end;
end;
%=== end ===
It seems to work as the scaling in the 'affine' approach are 0, 1, 1,
and 1 as they should be, but as always, there is no guarantee that any
of this works as intended. Also, in how far this will impact your final
segmentation results is another question, but one that may more easily
be estimated given the results of these simulations.
Cheers,
Marko
YAN Chao-Gan wrote:
> Hi Yuzheng,
>
> Actually, I have no experience for pediatric template yet. I guess 'no'
> might be a better option for your case.
>
> However, I am curious for recommendations from other researchers here.
>
> Thanks,
>
> Chao-Gan
>
> On Thu, Dec 11, 2014 at 10:25 AM, Yuzheng Leeson Hu <[log in to unmask]
> <mailto:[log in to unmask]>> wrote:
>
> Thanks Chao-Gan. It definitely helps a lot. But I still have a
> following up question about the impact of different "Affine
> Regularisation" on the final segmentation result. By searching
> online, I find previous discussion on this topic
> (https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0905&L=spm&D=0&P=84359).
> Marko suggested that it was "not that much". But I still don't know
> "how much" it is and have no idea which option for affine
> regularisation should be chosen if a pediatric TPM template is
> used. Do you have any suggestion or comment on this? Thank you.
>
> Best,
>
> Yuzheng
>
> On Wed, Dec 10, 2014 at 11:09 PM, YAN Chao-Gan <[log in to unmask]
> <mailto:[log in to unmask]>> wrote:
>
> Hi Yuzheng,
>
> Thank you very much for your interest in our work!
>
> To use the pediatric TMP files you have generated, you can:
> 1. Replace the file of SPMPath/toolbox/Seg/TPM.nii. Then
> everything else is the same in the GUI.
> 2. Or you can change the code in DPARSFA_run.m Line 1959
> SPMJOB.matlabbatch{1,1}.spm.tools.preproc8.tissue(1,T1ImgSegmentDirectoryNameue).tpm{1,1}=[SPMPath,filesep,'toolbox',filesep,'Seg',filesep,'TPM.nii',',',num2str(T1ImgSegmentDirectoryNameue)];
>
>
> If you want to use 'no' for "Affine Regularisation", I will
> suggest you to do the following:
> 1. Setup all the parameters in the GUI.
> 2. Click "Save" to save the parameters in a .mat file.
> 3. Load the .mat file.
> 4. Input in command
> line: Cfg.Segment.AffineRegularisationInSegmentation = 'no';
> 5. Input in command line: DPARSFA_run(Cfg); %This is the command
> line way for running DPARSF.
>
> Hope this helps.
>
> Best,
>
> Chao-Gan
>
>
> On Wed, Dec 10, 2014 at 10:20 PM, Yuzheng Leeson Hu
> <[log in to unmask] <mailto:[log in to unmask]>> wrote:
>
> Hi Chao-gan and other SPMers,
>
> I find DPARSF is easy to use, and appropriate Chao-gan and
> his colleagues' effort to develop and maintain such a good
> toolbox. Here I have a question about the feasibility of
> using pediatric template for normalization with "New
> Segmentation +DARTEL" in DPARSF. With TOM8 I have generated
> 6 pediatric TMP files which can be used in the "New
> Segmentation". However, I am confused by options for "Affine
> Regularisation". There are four options in SPM8,
> "no","mni","eastern" and "subj". But there are only two
> options available in DPARSF, "mni" and "eastern". What
> shall I do with these options to make sure the children's
> brain will not be stretched to the size of adult?
>
> Thank you.
>
> Yuzheng
>
> --
>
> */Yuzheng Hu Ph.D./*
>
> /Postdoc-Fellow
> /
> /National Institute on Drug Abuse (NIDA),/
> /National Institutes of Health (NIH),/
> /BRC 07A505, 251 Bayview BLVD,Baltimore, MD 21224, USA/
> /Phone +1-443-613-1206 <tel:%2B1-443-613-1206>/
>
>
>
>
>
> --
>
> */Yuzheng Hu Ph.D./*
>
> /Postdoc-Fellow
> /
> /National Institute on Drug Abuse (NIDA),/
> /National Institutes of Health (NIH),/
> /BRC 07A505, 251 Bayview BLVD,Baltimore, MD 21224, USA/
> /Phone +1-443-613-1206 <tel:%2B1-443-613-1206>/
>
>
--
____________________________________________________
PD Dr. med. Marko Wilke
Facharzt für Kinder- und Jugendmedizin
Leiter, Experimentelle Pädiatrische Neurobildgebung
Universitäts-Kinderklinik
Abt. III (Neuropädiatrie)
Marko Wilke, MD, PhD
Pediatrician
Head, Experimental Pediatric Neuroimaging
University Children's Hospital
Dept. III (Pediatric Neurology)
Hoppe-Seyler-Str. 1
D - 72076 Tübingen, Germany
Tel. +49 7071 29-83416
Fax +49 7071 29-5473
[log in to unmask]
http://www.medizin.uni-tuebingen.de/kinder/epn/
____________________________________________________
|
