Hi,
You are correct that PNM is likely to work better when working from the original data rather than the output of ICA.
We do not normally spatially smooth the PNM regressors, but you could if you wanted to, although be wary of doing this through-slice as there can be substantial differences there, based on the slice acquisition timing.
Using a peak waveform for the cardiac trace is fine in PNM, and is explicitly catered for with the "Pulse Ox Triggers" button, since this is such a common trace to work with.
All the best,
Mark
On 22 Oct 2013, at 06:32, Tae-Ho Lee <[log in to unmask]> wrote:
> Thanks a lot! It was really helpful to have an idea for preprocessing.
> I have several additional questions based on your comments. I would appreciate your comments.
>
> 1. If I run both ICA-denoising and PNM, then which step should be the first. I simply guess that ICA-denoising should be first because the results of physio regressor creation would be more precise (?) when they are created based on the denoised EPI. Is my understanding correct?
>
> 2. I'm also curious about spatial smoothing for PNM regressors. I guess it would be okay as long as the ICA is okay with PNM, right?
>
> 3. Is it okay that can I put cardiac data into the PNM function as a binary form (peak = 1, no peak = 0) instead of whole cardiac wave. The reason is that peak detection of PNM always requires additional visual inspection, so I think that putting peak information (based on R-R interval) is much easier in terms of visual inspection for possible correcting step.
>
> Best,
> Tae-Ho
|