I would use both tools, as ICA-based denoising tends to be better for spatially specific artifacts whereas global artifacts are better removed using externally derived regressors like motion regressors or PNM. What we do in the HCP is to run ICA on the original data and classify the components into signal and noise. Then we regress out 24 motion parameters from both the timeseries data and the noise component timecourses. Finally we regress the resulting noise component timecourses from the resulting timeseries data. If you were to use PNM (which we aren't currently using at this time), you would add those regressors in at the same stage as the motion parameters.
Matt.
Hello FSL listmembers,
We collected pulse data during our FMRI scans in order to facilitate removal of cardiac noise from the BOLD signal. My tentative plan is to use the Physiological Noise Model tool for this purpose. However, assuming that a model-free solution (MELODIC) would still be useful to remove non-cardiac noise, the question is this: In which order should we apply these two tools?
One option: Run PNM on the original data to create the PNM EVs. Then run MELODIC on the original data to identify all noise components; denoise based on the identified MELODIC components. Then run FEAT on the MELODIC-denoised data with the PNM regressors as EVs at the first level.
The worry with this option is that MELODIC will result in removal of most/all of the noise, including the cardiac noise. If that's true, the PNM regressors at the first level won't "do" much and may render less precise parameter estimates. In that case, should I plan to only use MELODIC? Or maybe run PNM first to create cardiac-denoised data that then gets pushed through MELODIC? Any thoughts are welcome. Thanks in advance.
Best,
Heather
P.S. I searched the FSL website and mail archive and didn't find info addressing this issue. And the PNM user guide is not currently available at
http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/PNM/UserGuide. Apologies if my questions were addressed elsewhere and I missed it.