Hi everyone,

I am recently starting to work with a very interesting dataset of quite sensitive patients and unfortunately due to complications with having these patients in the scanner, physiological data was not always reliably acquired in all patients. 

I am looking for advice in the preprocessing phase. Right now I have a very robust pipeline that relies at a point on the physiological acquisition, however for the reasons mentioned above this won't work in every patient. I could used ICA-based denoising, but I am not confident as it seems to underperform correction with Retroicor. I was wondering if anyone has experience with this and would lend some advice. I have thought about using ICA-based only on the patients where data is missing and then using a covariate later to model the slightly different approach, but am not sure if that is good enough to proceed. Would be a shame to discard some of these patients as they are not easy to acquire.

Best,

Herberto