We tried CAT12 and we adopted it, Thanks.
Hervé
-----Message d'origine-----
De : Christian Gaser [mailto:[log in to unmask]]
Envoyé : vendredi 17 mars 2017 09:31
À : [log in to unmask]; LEMAITRE Hervé Université Paris Sud
Cc : Christian Gaser
Objet : Re: VBM8 segmentation
Dear Hervé,
I have tried to segment that data set and obtained more reliable results by decreasing the number of "Gaussians per class" to "1 1 2 3 4 2". The first two classes are GM/WM and to only use one gaussian per class sometimes helps for difficult data.
However, I had some difficulties with skull stripping which was too aggressive. Thus, you have to play a little bit with the "Gaussian per class" option for GM/WM and also with skull-stripping.
I guess that you are still using VBM8 for compatibility reasons because of some existing data that were processed with VBM8. However, have you ever tried CAT12 for that data set?
Best,
Christian
On Thu, 16 Mar 2017 12:14:46 +0000, Herve Lemaitre <[log in to unmask]> wrote:
>Dear all,
>
>We are trying to segment a T1 image of an 11 year old child with the VBM8 toolbox in SPM8. Some parts of the brain that appear clearly as grey matter on the T1 images are not segmented as grey matter but as csf (see cross hair). We use the standard options except we use age specific TPM from TOM8. We tried with different other options but without success. Is there a way of tuning vbm8 to get this right?
>Thanks,
>Hervé
>
>PS: you can get the original image here https://transfert.u-psud.fr/z3l9
>
>
|