Dear Martin,
>I want to ask you if the procedure of estimating hemodynamic
>parameters by using Volterra kernels - as described in your papers
>[1,2] is somewhere available in Matlab code.
>I need to perform estimation of 6 hemodynamic parameters from voxel
>time series in order to build a priors for later evaluation of
>hemodynamic state-space model by using nonlinear Kalman filter. I am
>already done with nonlinear estimation of states but still fighting
>with parameter estimation part. Unfortunately, I am not familiar
>with application of Volterra kernels (but willing to learn).
>Therefore, I would be very grateful for any help from you. Thank you
>beforehand.
There is no need to use Volterra kernels - you can estimate the
parameters of a hemodynamic model directly
using the same scheme employed by DCM. Analyze your data in the usual
way in SPM8 and then display the
results. At the bottom of the interactive window is a button called
'hemodynamics'. If you press this and
answer the questions you should get estimates of the parameters like
those shown below.
The variable
>> Ep =
0.5667
0.2620
2.0207
0.3268
0.3767
-0.5412
0.1109
0.6263
0.0360
in working memory contains the posterior expectation (MAP) estimates.
These are described in
>> help spm_hdm_priors
returns priors for a hemodynamic dynamic causal model
FORMAT [pE,pC] = spm_hdm_priors(m,[h])
m - number of inputs
h - number of hemodynamic modes (default = 3)
pE - prior expectations
pC - prior covariances
(5) biophysical parameters
P(1) - signal decay d(ds/dt)/ds)
P(2) - autoregulation d(ds/dt)/df)
P(3) - transit time (t0)
P(4) - exponent for Fout(v) (alpha)
P(5) - resting oxygen extraction (E0)
P(6) - ratio of intra- to extra- (epsilon)
vascular components of the
gradient echo signal
plus (m) efficacy priors
P(7) - ....
I hope this helps - Karl
PS If the scheme breaks for very large BOLD signals (greater than
4%), increase V0 in spm_gx_hdm until the
scheme convergences.
>[1] K.J. Friston et al.: Nolinear Response in fMRI: The Ballon
>Model, Volterra kernels, and other hemodynamics, (1998).
>[2] K.J. Friston: Bayesian estimation of dynamical systems: An
>application to fMRI, (2002).
Emacs!
