 |
 |
On Mon, 17 May 2010 17:20:01 +0200, michel grothe <[log in to unmask]> wrote: Dear Michel, > >Dear Mr. Gaser, > >thank you very much for your fast reply. However, I guess that I have to > reformulate my second question: > > > >How are the jacobian maps computed? As far as I understand the >registration in VBM8 it involves an initial affine registration to MNI >space and subsequent Dartel warping of the GM+WM segments (in MNI space) > to the IXI-DARTEL-Templates (in MNI space). This would mean that the >Dartel-flow-fields just model remaining non-linear effects. Are the >Jacobians derived from these Dartel-flow-fields (which would mean that >global head size is already accounted for in the Jacobian maps, much >like the m0-modulation) or are they derived from a composition of the >initial affine registration and the Dartel-warps? In the latter case the > jacobians would represent volumetric changes from native space to the >IXI-Dartel-template in MNI space, similar to modulated >(affine+non-linear) tissue segments. Given that the VBM8-GUI just gives >the options "none" and "modulated" for Jacobian maps I�m a bit confused >about the "modulation" of the Jacobians. > The jacobians are estimated using only the non-linear terms of the spatial normalization. No additional composition with the affine term is applied. The composition with the affine term is only considered if you save modulated image and the option "affine + non-linear" is selected. This is different from John's approach, but as you already assumed it accounts for different brain/head sizes and simplifies the statistical analysis. It is equivalent to a proportional scaling with the (inverse) scaling factor of the affine transformation in the statistical analysis. Regards, Christian ____________________________________________________________________________ Christian Gaser, Ph.D. Assistant Professor of Computational Neuroscience Department of Psychiatry Friedrich-Schiller-University of Jena Jahnstrasse 3, D-07743 Jena, Germany Tel: ++49-3641-934752 Fax: ++49-3641-934755 e-mail: [log in to unmask] http://dbm.neuro.uni-jena.de > > > >Best regards, > >Michel > >> Date: Mon, 17 May 2010 06:56:14 +0100 >> From: [log in to unmask] >> Subject: Re: Beta Version of VBM8 Toolbox >> To: [log in to unmask]; [log in to unmask] >> CC: [log in to unmask] >> >> On Mon, 17 May 2010 00:39:28 +0200, michel grothe <[log in to unmask]> wrote: >> >> > >> >Dear Mr. Gaser, >> > >> >I have some additional questions on the new VBM8 Toolbox. >> > >> >1. Which tissue-types are used for the Dartel-normalisation? >> GM + WM, same as standard Dartel approach >> >> > >> >2. How are the Jacobians modulated? Am I right in assuming that the >> >initial affine transformation to MNI space is used for that >> >normalisation and hence does not involve non-linear components? If so, >> >is there a possibility to get unmodulated jacobians in MNI space, so >> >that global head size is already accounted for? >> Check my website for more information: >> http://dbm.neuro.uni-jena.de/vbm/segmentation/modulation/ >> >> If you want modulated images, where global head size is corrected for you can use the modulation for non-linear effects only (which is the default). This will be indicated by a prepending "m0" instead of "m". The advantage of VBM8 Dartel integration is that no additional steps for MNI normalization and corrections are necessary. >> >> Regards, >> >> Christian >> >> ____________________________________________________________________________ >> >> Christian Gaser, Ph.D. >> Assistant Professor of Computational Neuroscience >> Department of Psychiatry >> Friedrich-Schiller-University of Jena >> Jahnstrasse 3, D-07743 Jena, Germany >> Tel: ++49-3641-934752 Fax: ++49-3641-934755 >> e-mail: [log in to unmask] >> http://dbm.neuro.uni-jena.de >> >> >> > >> > >> > >> >Thanks for this very helpful toolbox. >> > >> > >> > >> >Best regards, >> > >> >Michel Grothe >> > >> >> Date: Sun, 16 May 2010 23:15:18 +0100 >> >> From: [log in to unmask] >> >> Subject: Re: [SPM] Beta Version of VBM8 Toolbox >> >> To: [log in to unmask] >> >> >> >> Dear Paul, >> >> >> >> > Thanks for your toolbox. I have some questions about parameter setting >> >> > >> >> > 1. How to change DARTEL export image voxel size, i think the default setting about DARTEL export image is 1.5 cubic >> >> The voxel size is always set to the voxel size of the tissue probability map (TPM) image. However, if you don't need a customized Dartel template you can use the Dartel approach, which is integrated in VBM8. This Dartel approach will use the Dartel MNI template provided with VBM8 and no additional export or Dartel use is neeeded. >> >> >> >> > >> >> > 2. How to change HMRF weighting ? >> >> > (In VBM5, we can choose different HMRF weighting in GUI. If i want to use iterative HMRF weighting, how can we choose in this version?) >> >> In VBM8 two de-noising techniques are used. The first is a non-local means filter as preprocessing step. The filter size is automatically estimated based on the local variance of your image. The second is a MRF, which is integrated in the segmentation process. Again the weighting is now internally estimated based on the noise in your image. However, both approaches can be weighted. The non-local means filter is by default weighted by 0.7, which shows best performance for segmentation. The weighting for the MRF is fixed to 0.15. Both values are optimal for a large variety of data and usually it is not necessary to change the weighting. The weighting can be rather used to deselect the filter by setting the weighting to 0. Anyway, I have now added the option to change both weightings. This will also solve some problems with batch files, where changes of these parameters in cg_vbm8_defaults.m were not considered. Thanks to Jose V. Manjon for pointing to that problem. >> >> The new version can be downloaded at: >> >> http://dbm.neuro.uni-jena.de/vbm8/vbm8_r331.zip >> >> >> >> Or simply use the update function in VBM8. >> >> >> >> Regards, >> >> >> >> Christian >> >> >> >> ____________________________________________________________________________ >> >> >> >> Christian Gaser, Ph.D. >> >> Assistant Professor of Computational Neuroscience >> >> Department of Psychiatry >> >> Friedrich-Schiller-University of Jena >> >> Jahnstrasse 3, D-07743 Jena, Germany >> >> Tel: ++49-3641-934752 Fax: ++49-3641-934755 >> >> e-mail: [log in to unmask] >> >> http://dbm.neuro.uni-jena.de >> >> >> >> > >> >> > Best >> >> > >> >> > Paul >> > >> >_________________________________________________________________ >> >http://redirect.gimas.net/?n=M1004xSkyDrive2 >> >Ihre Daten brauchen Platz? SkyDrive gibt Ihnen 25 GB - gratis! >> >> > >_________________________________________________________________ >http://redirect.gimas.net/?n=M1004xWin72 >Windows 7 - Alles was Du brauchst und noch viel mehr!