Dear Michel,
On Fri, 13 Dec 2013 08:13:59 +0100, michel grothe <[log in to unmask]> wrote:
>Dear Christian,
>
>thank you very much for the clarification.
>Just one follow-up question:
>
>> Only GM and WM are used for normalization. All other classes are only used as prior for the segmentation.
>
>It was my understanding that the segmentation in VBM8 is purely intensity-based and does not use segmentation information from tissue-priors. Does that mean that specifying "SPM default low-dimensional normalization" uses both the segmentation and normalization algorithm from "New segment", as these two are not entirely separable in the unified segmentation approach?
Sorry, my fault. You are right that the segmentation approach in VBM8 is working without any priors. The inital SPM segmentation (New segment) is only used for removing non-brain parts of the T1 image (skull-stripping).
>Still, the native tissue maps (p*) outputted in either (normalization) case come from the VBM8-specific segmentation algorithm and should be identical, right?
Yes, the native segmentations are independent from the type of spatial normalization in VBM8.
Best,
Christian
>
>Thanks for your help.
>
>Best,
>Michel
>
>
>> Date: Thu, 12 Dec 2013 21:14:43 +0000
>> From: [log in to unmask]
>> Subject: Re: [SPM] Normalization in VBM8 toolbox
>> To: [log in to unmask]
>>
>> Dear Michel,
>>
>> On Wed, 11 Dec 2013 14:12:08 +0000, Michel Grothe <[log in to unmask]> wrote:
>>
>> >Dear Christian (or anyone else eager to answer ;-)),
>> >
>> >I was wondering which kind of SPM "standard" normalisation the VBM8 toolbox uses when changing the normalization settings from "high-dimensional DARTEL" to "SPM low-dimensional"? Old-fashioned "Normalise"-Button or the one in SPM5-style unified segmentation?
>> It's the normalization that is used in the unified segmentation approach ("New segment") in SPM8.
>>
>> >
>> >I noticed that the TPM-file where images are normalised to in the case of "low-dimensional" normalisation is taken from the "New segment" toolbox, meaning that it includes a total of 6 tissue-types. Are all parts of this TPM-file used for matching the tissue-segments, or just the GM and WM (and CSF?) partitions.
>> Only GM and WM are used for normalization. All other classes are only used as prior for the segmentation.
>>
>> If I wanted to include my own TPMs, would it be sufficient to specify a TPM-file with just the 2 volumes for GM and WM? The manual recommends using the TOM toolbox, but I already have mean GM and WM maps of an independent study population (actually a DARTEL template), which I would like to use.
>> No, this will not work, because the priors for segmentation are necessary.
>>
>> >Similarly, if I wanted to compare directly the outcomes from "high-dimensional" vs "low-dimensional" normalization strategies, I would like to ensure that the reference template is identical for the two approaches. Thus, would it be reasonable to replace the default TPM-file with the GM and WM partitions from the DARTEL-IXI-template (_6), or would you recommend a different approach? I know, the TPM-file and the IXI segments should be analogue because both are in MNI space, but still...
>> Because the spatial normalization is working so different a direct comparison using the same template will not work. The standard normalization (from new segment) will use GM/WM of one template (TPM.nii). In contrast Dartel needs templates for each iteration step with increasing spatial accuracy. The Dartel-template that is provided with VBM8 (or for any other template that you obtain after creating a customized template using Dartel) contains of 2 tissues classes (GM/WM) and 6 iteration steps.
>>
>> Best,
>>
>> Christian
>>
>> >
>> >Thanks in advance for your help.
>> >
>> >Best,
>> >Michel
>> >
>
