Dear Amir and other experts:
I spent some time in DCM with neural mass models (NMM) for resting-state fMRI, but was not able to figure it out. Based on my understanding, DCM for resting state fMRI can be estimated using either stochastic DCM or CSD approach. In SPM12, the core inversion routine for DCM with NMM is spm_dcm_fmri_nmm.m. For stochastic DCM, if I set the variable DCM.options.stochastic to 1 in the routine, can it be applied to resting-state fMRI data? I think it may require more work than that.
For the CSD option, I know the core inversion routine is spm_dcm_fmri_csd.m, but I am not sure how to combine spm_dcm_fmri_nmm.m with spm_dcm_fmri_csd.m together. The main question is how to specify the model. Under spm_dcm_fmri_nmm.m, the first two lines for model specification are:
M.IS = @spm_gen_fmri; % integration scheme
M.fn = fn; % neuronal equations of motion
Under spm_dcm_fmri_csd.m,
DCM.M.IS = 'spm_csd_fmri_mtf';
DCM.M.FS = 'spm_fs_fmri_csd';
DCM.M.g = @spm_gx_fmri;
DCM.M.f = @spm_fx_fmri;
That means the integration scheme is different and spm_gen_fmri is made specifically for NMM. Also, under spm_dcm_fmri_csd.m, the spectral responses need to be calculated first before calling the function spm_nlsi_GN for model inversion.
It would be greatly appreciated if you can give me some ideals or suggestions on how to modify the routine spm_dcm_fmri_nmm.m for resting-state fMRI data. Thank you very much!
Best regards,
Guoshi Li