Best interpolation for RR intervals

Hello, i was wondering which is the best kind of interpolation to obtain a good tachogram from the RR intervals. I have tried:

 y = interp1(t,x,t_out,'spline');
 y = interp1(t,x,t_out,'cubic'); 
 y = csaps(t,x,0.5,t_out);  %0.5 randomly selected between 0-1
 y = csapi(t,x,t_out);

But i am not sure i understand the difference between them. Especially between interp1 with "spline" argument and csapi. Thank you in advance for your responses.

Answer 1

[Star Strider]

R-R intervals normally vary for a number of reasons. I would simply do a linear interpolation if you want them all to have the same time base. Retaining the heart rate variability is important, and a linear interpolation is most likely to preserve it.

[George Lazaridis]

it's very logical what you say, but linear is not really recommended anywhere in literature.

[Star Strider]

I just looked up and read (and attached) what appears to me to be a good review on the issue, and it didn’t express any particular preference. (Found on a PubMed search on ‘("R-R interval" & interpolation) & english[la]’.)

Unless there is a physiological reason to assume a non-linear trend (that I am not aware of), linear seems to me to assume less than other methods. There does not seem to me to be a specific reason not to use it in preference to other methods. Generally, the Hering-Breuer reflex, preload changes, and other variables operate at sufficiently low frequencies that nothing would be lost using a linear approach. Any sort of interpolation may assume trends that do not actually exist, but linear seems to me to assume less than nonlinear methods. I’m not necessarily defending it (I haven’t done HRV studies in a while and will likely not ever do them again), but I see no reason to reject it in favor of others.

I don’t have the Curve Fitting Toolbox (Statistics and Optimization instead) so I can’t comment on csaps and csapi.

[George Lazaridis]

Hm, in the majority of papers in HRV analysis they use spline interpolation. I don't know why.

[Star Strider]

And the papers themselves likely didn’t explain their choice of spline interpolation.

I don’t have access to ‘Effect of missing RR-interval data on nonlinear heart rate variability analysis’, but they apparently compared different interpolation and analysis methods. It might be worth looking this up to see what their results were. (I would not object to your uploading it here.) From the abstract, it seems the sort of study I would do had I the resources.