Hi Michael,
What you suggest to get your regressors looks okay to me, but you will have
to do a final downsampling step such that the number of elements in your
regressor matches the number of image volumes you use in your model (e.g
using linspace to find the indices of convraw to use). Finally, make sure
you have got your offset w.r.t. to the images right.
Regarding the 2nd part of your questions: MR physicists keep telling me that
a TR of less than 2 sec will eventually lead to worse signal as we to accept
the physical properties of tissue. However, my feeling is that you very much
oversample your physiological variable anyway, and that a TR of 2 will
probably capture its fluctuation quite well - especially after you have
convolved it with the HRF. Out of interest - what is it that is sampled at
60 Hz? The only physiological measure I can think of which fluctuates up to
that speed is neuronal (or muscle) activity...
Hope this helps,
Helmut
----- Original Message -----
From: "michael rogan" <[log in to unmask]>
Sent: Tuesday, May 16, 2006 7:35 PM
Subject: Help: convolving high resolution physiological regressor with hrf
>I have a continuous physiological measure (raw_data), 60 measurements/sec,
> taken during scanning with a TR=2.
>
> It seems to me that I should convolve the raw_data with hrf prior to
> binning
> in 2sec bins.
>
> I have noted the following procedure in other posts:
> hrf=spm_hrf(TR)
> convraw=conv(raw_data, hrf);
> convraw=convraw(1:length(raw_data));
>
> Is this the correct method for this data? Should the temporal resolution
> of
> the hrf be closer to the temporal resolution of the raw_data?
>
> THanks for any suggestions.
>
|