Dear Zoe,
On Tue, 17 Jul 2012 11:56:32 +0000, Hyde, Zoe <[log in to unmask]> wrote:
>Dear Christian and Marko,
>
>Thanks very much to both of you for your replies to our previous question.
>
>Christian, I hope you don't mind a couple of follow-up questions to clarify the pipeline, especially as yours and Marko's replies differed regarding MNI space. First, we were surprised to find (if we're reading your response correctly) that choosing 'DARTEL export - affine' for the initial segmentation using 'Estimate and Write' in VBM8 means that the segments are written in MNI space. Could you confirm whether this is indeed the case? The instructions and the manual don't mention this as a way of normalising into MNI space, so we would like to be absolutely sure before proceeding. Is this due to the fact that we used the default DARTEL template which is in MNI space as mentioned in the VBM8 manual (p.29)? And does it not make a difference that we have used our own custom TPMs?
Affine normalization also includes scaling to MNI space, that is separately applied in the Dartel toolbox. Thus, if you use the "export affine" option to write segmentations you don't need to additionally transform your Dartel results to MNI space. Furthermore, the Dartel template provided with VBM8 is in MNI space (as the name implies).
The custom TPM image is only used for the initial spatial normalization, but also influences the estimation of the affine transformation and therefore the exported affine transformed segmentations. Thus, it is especially recommended to use a customized TPM for children (via the Template-O-Matic toolbox).
>
>Secondly, during this step, does it matter whether we choose the low dimensional or high dimensional option under 'Spatial normalization'. Marko's reply suggested that it wouldn't matter since we aren't actually normalising the data at this point in the pipeline, and that we should therefore use the low dimensional option for speed. But if the affine transformation is also normalising to MNI space, then perhaps it does matter after all.
The exported affine transformed segmentations are unaffected of the subsequent non-linear normalization. Thus, you can use the low-dimensional normalization to save time.
>
>Presumably after this step, we then use the 'DARTEL export - affine' segments to create our DARTEL template as we mentioned in our previous post. But previously we had thought the template would be written in study-specific space. Are you in fact saying that the template will be in MNI space?
As I mentioned above, affine transformation also scales images to MNI space...
>
>Finally, as a final step you suggest using 'the Dartel toolbox to apply the warpings with modulation but without additional MNI normalization.' Would this be using the 'Normalise to MNI space' option in DARTEL? If so, did you mean that we do not need to put the template in at this stage because the flow fields will already be what is required to transform our images into MNI space? And are you saying that once we've used the VBM8 toolbox for initial segmentation, that we then don't use it again during pre-processing? Our original plan was the following: After having created our DARTEL template, we were going to segment our original T1 scans using the Estimate and Write from VBM8 with our customized tissue probability maps and the customised DARTEL template. However, based on your email it seems that if VBM8 was used in the initial segmentation (to generate the DARTEL export - affine), there is no additional benefit of using it thereafter. Our rationale for using VBM8 at the final stage was due the additional cleaning options offered by VBM8 (i.e. MAP, PVE, SANLM and MRF) and the possibility of getting the normalised bias corrected T1 scan to create a mean T1 to overlay our results.
The flow fields can be applied to you affine transformed segmentations. In this case you can use the potential benefits of the VBM8 toolbox.
I hope that it's clearer now...
Best,
Christian
>
>Thanks very much in advance for your advice.
>
>Best wishes,
>
>Zoe
>
>Zoe Hyde
>Research Assistant
>
>Developmental Risk and Resilience Unit
>Clinical, Educational and Health Psychology
>Division of Psychology and Language Sciences
>
>University College London
>26 Bedford Way
>London WC1H 0AP
>
>[log in to unmask]<mailto:[log in to unmask]>
>Tel: +44(0)20 7679 5322<tel:%2B44%280%2920%207679%205322> (internal: 25322)
>http://www.drru-research.org/
>
>
|