Hi Catie,
I hope this helps....
This 32 offset value comes from spm_get_ons.m.
The stimulus function or box-car function has a length of nscans *
microtime resolution (usually the number of slices) + 32.
If you plot the field u in the U structure, you will realize that there
is an offset of 32.
just do figure, plot(U.u) and zoom
This value of 32 is the length of the canonical hrf in seconds, which is
computed in spm_get_bf.
Also note that the length of bf is 32 * TR/microtime resolution, where
TR/microtime resolution is your time between two matlab points.
In the case you are not using the any of the canonical hrf choices (hrf
or hrf + temporal derivative or hrf + temporal and dispersion
derivative) then the default value of the window for your basis
functions is 32 (see spm_get_bf.m), unless you input another value. In
that case, I am not really sure whether this 32 offset is still
appropiate or it must be changed to SPM.xBF.length.
Then, your set of basis functions (bf) is convolved in spm_volterra with
your stimulus function (U.u), and the output of convolution has a length
of length(U.u) + length(bf) -1, so you must pad zeros before your
stimulus function. According to matlab, there is no need to pad zeros to
your signal since convolution is done in frequency domain. But i am not
really sure if this padding is crucial for some some other spm files
when plotting results. Any suggestions?
Then downsampling is carried out and these signals in X are used to
build the design matrix. So in my opinion, this is not a minor question
since the downsampling is crucial for event related fmri and, especially
when your TR is large. Note that there is another offset of T0 which is
your microtime onset, and according to the SPM5 manual this must be set
equal to the time point of your reference slice. So this is clearly
related to the slice timing problem.
Cesar Caballero
Phd Student
Computer Science School/SPMMRC
University of Nottingham
Catie Chang wrote:
> Hi -
>
> This is a very very minor question, but I was just noticing that in
> spm_fMRI_design.m (in SPM5), a 32-bin offset is used when resampling
> the regressors at acquisition times. I was wondering where the 32
> comes from.
>
> Thanks :)
> catie
This message has been checked for viruses but the contents of an attachment
may still contain software viruses, which could damage your computer system:
you are advised to perform your own checks. Email communications with the
University of Nottingham may be monitored as permitted by UK legislation.
|