Print

---

Dear Jiang:

The SPM5 unified segmentation spontaneously warp the subject to ICBM 
template, hence it is kind of warping GM with GM template, WM with WM 
template, CSF with CSF template. And John said it should be more 
accuracate, as it is closer to the practice of optimized VBM.
If you normlaize your functional images with a EPI images, it will be 
based on mean square differences. Hence the penalty of mis-matching will 
be different. However, better normalization with the *_seg_sn.mat is 
based on the assumption that you good co-registration of your fMRI with 
structural MRI. If your EPI has strong distortion due to whatever 
reason, the normalization may not work that well as the parameters are 
calculated from the normalization of the structural MRI.

Best,
Carlton
jiang wrote:

> Dear SPMers,
>
> 1) I have a basic question about normalisation :
>
> with the examples given in the spm5 manuel, the normalisation 
> procedures are : segmentation the structural image, use the 
> segementation parameters to normalise functional images ;
>
> I have learnt to do like the following :  estimate and normalise the 
> functional images to EPI tempelet, use the parameters obtained to 
> normalise structural image (without segmentation).
>
> What are the  differences between these procedures, which procedure is 
> more pertinent in the purpose to localise the functional responses ?
>
> 2) at the step of the first level analysis, can I define in SPM a 
> comparison between the two sessions of the same subject like define 
> the factors for the different conditions of one session ?
>
> thak you very much !
> Tao JIANG
>
>