Dear Philippe,

Many thanks for your interest in our work and my apologies for the delayed reply - I am currently very busy.

As a (very brief) response to your questions:

(1) This is a misunderstanding - the model does refer to dS/S0.  Please note that all hemodynamic state variables are normalized with regard to their values at rest and that generally, both the neural and hemodynamic model components of DCM consider signal deviations from rest (note the role of the Taylor expansion around rest in both cases).

(2) The implementation in SPM uses the hemodynamic output equation derived in the appendix to Stephan et al. 2007 (NeuroImage 38:387-401).  You may find this derivation helpful to see why the factors affecting k2 in our model differ from those by Obata et al. 2004 (Neuroimage 21:144-153).

Best wishes,
Klaas



Von: Philippe Pouliot <[log in to unmask]>
An: [log in to unmask]; [log in to unmask]
Gesendet: 15:41 Montag, 21.November 2011
Betreff: Obata 2004 vs. Stephan 2007
Dear Dr. Stephan,

We are wondering about 2 points regarding your nice paper
Comparing hemodynamic models with DCM, NeuroImage 38, 2007.

1) In SPM8, the data fed to spm_gx_hdm (or spm_gx_dcm or spm_gx_fmri)
is the Bold signal S (called Y in the code), but we believe it should
be Delta S/S0.
Is this a mistake in the implementation of the measure model in SPM8?
Namely that the calculation of Delta S/S0 (with an estimation of what
S0 is) is missing
from the code?

2) In Obata et al. 2004, NeuroImage 21, 144, in the Appendix, equation
A12: there is
a factor of v following k2, but this factor is missing from the SPM8
implementation, and
also in various equations in your article (equation 2 and 13). This
makes no difference
to the linear version, but the nonlinear versions appear to be
different. Please could
you let us know if this omission was intentional or not?
Many thanks for your help,

Philippe Pouliot, Ph.D.
Montreal Heart Institute
Ecole Polytechnique Montreal