Quality of cardiac recordings

[HDhanis]

Hi Lars,

I hope you're doing well! I was wondering if I could have your input on how to assess the quality of the cardiac recordings of a dataset I am about to analyse.

For context: for this dataset both cardiac and breathing signals were recorded, however (and very unfortunately), something must have gone wrong in the configuration files during acquisition and instead of having the pulse oxymeter signal available I only have the automatic R-peak detection. The problem is that these acquisitions are of patients with Parkinson's disease and the quality of the cardiac signal varies considerably. For some the intervals between the peaks are as expected and reflect a regular heartbeat, but for others the signal is really bad, oscillating between way too close peaks (potentially due to tremors) and absence of peaks (likely due to cold extremities realted to autonomic dysfunction).

I see that in the figures put out by PhysIO I get sometimes a warning on figure 4 (diagnostics) saying "There seem to be wrongly detected heartbeats in the sequences of pulses (...)"/. And in the output physio files I can see this on the verbose field or otherwise on the ons_secs -> c_outliers_high and c_outliers_low. Weirdly enough, outliers_high is never populated even though in the graph I can see scans that are past the upper limit. I would like to get some more objective metrics out of it, maybe PhysIO saves the interbeat intervals? Or other diagnostics I might have missed?

So my question is sort of an open question on your opinion of how to handle this. The dataset is large so I can lose some patients without it being a major issue, but say if it is 25% of patients I will need to implement some sort of ICA denoising rather than multiband-RETROICOR which is what I am doing and would like to keep doing.

Thanks in advance for your inputs.

Cheers,
Herberto