Dear Zhi,
You are right that y.X0 used by DCM is defined in spm_regions.m:
https://github.com/spm/spm12/blob/r7487/spm_regions.m#L206-L230
in part from SPM.xX.K.X0 defined here with spm_filter.m:
https://github.com/spm/spm12/blob/r7487/spm_fmri_spm_ui.m#L210-L217
As I mentioned before, it might end up being much simpler to perform the
surface-based GLM in SPM than trying to reuse the outputs from
FreeSurfer. Some adjustments in SPM might be required along the way but
it will be an improvement and will make it easier for the next person
wanting to perform a similar analysis to yours. In a previous email, you
mentioned that you tried but "the results were weird that only one slice
had data": could you elaborate? Are your data stored in the CIFTI format?
Best regards,
Guillaume.
On 28/02/2019 20:17, Zhi Li wrote:
> Hello SPM experts,
>
> I would like to apply DCM on surface-based mean time series. Given the image data were pre-processed and modeled with FreeSurfer, I cannot use SPM to extract times series or to get the X0 matrix directly from the SPM.mat. I know I could use "spm_filter.m" to get xY.X0, but how could I acquire Y.X0 or how to calculate it? I also found that in "spm_regions.m" the Y.X0 was directly taken from SPM.xX.xKXs.X(:,[SPM.xX.iB SPM.xX.iG]). It seems Y.X0 was generated in "model estimation", is it correct? I found that if I feed the 1st-level GLM in SPM with different images (e.g. smoothed vs non-smoothed images), the Y.X0 would be different.
>
> If the Y.X0 is not accessible without SPM, may I know that if Y.X0 influence the DCM modelling a lot or if it could be ignored? Any suggestions would be appreciated.
>
> Thanks and best withes,
>
> Lizhi
>
--
Guillaume Flandin, PhD
Wellcome Centre for Human Neuroimaging
UCL Queen Square Institute of Neurology
London WC1N 3BG
|