Yes, variability could arise at mulitple stages between the input
(experimentally-controlled stimulation) and output (BOLD signal), e.g, in
the stimulus-neural, neural-bloodflow, or bloodflow-BOLD mappings. Basis
functions cannot help here (only independent data). However, it is my
experience that in situations where the neural activity is pretty likely to
be short-lived (ie impulsive), the neural-BOLD mapping can probably be
characterised by about 3 dimensions (eg basis functions), such as SPM's
canonical HRF and temporal and dispersion derivatives (see, eg, multisubject
fMRI "random effects" demo on the SPM data website). If you are less sure
about the neural activity associated with your stimulations (eg whether it
is more sustained), you could use two sets of functions: one constrained set
for the neural-BOLD mapping (eg SPM's three basis functions), and a more
flexible (eg Fourier) set for the stimulus-neural mapping (as suggested, for
example, by Eric Zarahn in 2000, NI).
Rik
--------------------------------------------------------
DR RICHARD HENSON
MRC Cognition & Brain Sciences Unit
15 Chaucer Road, Cambridge,
CB2 2EF England
EMAIL: [log in to unmask]
URL: http://www.mrc-cbu.cam.ac.uk/~rik.henson
TEL +44 (0)1223 355 294 x522
FAX +44 (0)1223 359 062
MOB +44 (0)794 1377 345
--------------------------------------------------------
>-----Original Message-----
>From: SPM (Statistical Parametric Mapping)
>[mailto:[log in to unmask]] On Behalf Of Jason Steffener
>Sent: 18 April 2006 18:00
>To: [log in to unmask]
>Subject: [SPM] Basis functions modeling neuronal effects?
>
>
>Hello all,
>I have sent this post to both SPM and FSL so I
>apologize if anyone receives it twice.
>
>I was just thinking about the accurate use of basis
>functions to model fMRI time series data. They are
>used to model variance in the shape of the hemodynamic
>response function. However, what if the model of the
>neuronal response is not accurate, for some reason it
>doesn't accurately follow the stimulus paradigm. Then
>the basis set fit to the data will be accounting for
>the variance in the HRF AND the mismatch between the
>REAL neuronal response and what we assume it to be. So
>then the question is how do we tell what is HRF
>variance and what is neuronal variance?
>
>And if this is true. Then the approaches that apply
>restrictions on the basis function fits (Woolrich et
>al 2004, Friman et al. 2003, Calhoun et al. 2004,
>Worlsey et al. 2006) would need to be less
>restrictive? Or restrictive in a different way? And
>how does this idea affect results based on fitting a
>basis set to data and then determining the delay in
>the data or amplitude variations?
>
>I would love to hear all feedback and thoughts on this
>topic.
>
>Jason Steffener.
>
|