Hi,
well, you should use the one that gives you results that are, or seem
most appropriate for your population. Many people will not be surprised
if, in my opinion, using vbm8 (which uses priors only for normalization,
not for segmentation,)is preferable in the presence of a population not
well represented in the population contributing to the priors. In other
words, the less young-adult, healthy Canadians you have in your
population, the more it may be a good idea to move away from using a
segmentation approach using such priors :)
Cheers,
Marko
Andreas Hahn wrote:
> Hi Marko,
>
> Thank you very much for your support!
> Just to get back to the first question: which initial segmentation would
> be recommended as input for TOM (regardless of using affine or
> non-linear transformations)?
>
> Best,
> Andreas
>
> Marko Wilke schrieb:
>> Hi Andreas,
>>
>>> I would like to create a custom template for an elderly population (~76
>>> years) with the TOM8 toolbox.
>>
>> Not my usual population ;) but since I am not wholly innocent of
>> making TOM8 available, let me have a stab at this:
>>
>>> - Using SPM8's New Segment as TOM input, gives almost the same result as
>>> the already available VBM8 tissue prior (except for smoothness), which
>>> was somewhat surprising given the age of the sample.
>>
>> Well, to be fair they do not seem to be so similar, with some
>> differences apparent even on the screenshot in the temporal lobe and
>> the cerebellum.
>>
>>> - On the other hand, VBM8 as TOM input gives overall lower GMV values,
>>> which appears more reasonable for this sample (please correct me if I am
>>> wrong!).
>>
>> Can't tell for sure from the image you sent but I would tend to agree.
>>
>>> Still, for VBM8 the 3 background outputs are missing. Although
>>> Marko Wilke mentioned that code is available to get the additional
>>> outputs, I couldn't find it on the list - could someone please share
>>> this information?
>>
>> Well, some of it is on the list, and I have been hesitant to share
>> this again as it means fooling with Christian's code... so all bad
>> coming out of this is completely my fault and not Christian's (I still
>> assume no responsibility, though :)
>>
>> What you could try:
>> - edit cg_vbm_run at about l75 to specify outputs for the last 3
>> classes the same way as you specified for GM
>>
>> % now write class 4-6
>> for i=4:6
>> tissue(i).warped = tissue(1).warped;
>> tissue(i).native = tissue(1).native;
>> end
>>
>> - edit cg_vbm_write at about l448 to avoid clearing these last three
>> classes (simply commenting this out will do nicely)
>>
>> % DO NOT clear last 3 tissue classes to save memory
>> % for i=4:6
>> % cls{i} = [];
>> % end
>>
>> - edit cg_vbm_write at about l814 to loop over all 6, instead of three
>> classes, to write out all results (if you want results in native
>> space, the same applies to about l638).
>>
>> k1 = 1:6, % instead of 3
>>
>> This should (!) do it, but please check for yourself. You may also
>> want to check l412, l435, l868, and l884 where also not all classes
>> are "looped over", but I have not looked in detail what exactly is
>> happening there.
>>
>> Be aware that spm expects segmentation priors to sum up to 1 in every
>> voxel, so it is important to ensure that the maps match up before
>> passing stuff to TOM. TOM, when generating priors, will also do that
>> but larger discrepancies in the input data may screw up the following
>> calculations to begin with.
>>
>>> - Finally, when including non-linear transformations, should modulation
>>> be used?
>>
>> No. Priors reflect concentration, not volume.
>>
>>> Btw, could someone explain me the rationale for using only
>>> affine transformations with respect to template creation?
>>
>> Well, the jury is still out on what is better, but the rationale is
>> that you may want to keep the individual contribution of subjects to
>> the template, instead of making them all look the same before creating
>> a template (which would be possible, but besides the point). There are
>> differing opinions on this out there so please, voice them (if you
>> read this far ;)
>>
>> Hope this helps,
>> Marko
>>
>
--
____________________________________________________
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/
____________________________________________________
|