Comparison

[gencee]

Hey guys,
Congratulations on this library. It's very impressive and usefull. I am testing the peak detection on the MIT-BIH dataset.

It seems to be working well however the comparison function says the performance is low (close to 0 sensitivity in the case of beat 101). From the looks of the comparison function and the plots from the comparitor.plot, it seems everything becomes false as soon as there is one peak is missed or falsely detected.

[ChitandaXu]

I met the same problem.

[cx1111]

Thanks for noticing. I think it should be fixed, I tried on a couple of records like 101.

I've added the benchmark_mitdb function and tested xqrs using it:

• http://wfdb.readthedocs.io/en/latest/processing.html#wfdb.processing.benchmark_mitdb

import wfdb

from wfdb.processing import benchmark_mitdb, xqrs_detect
comparitors, spec, pp, fpr = benchmark_mitdb(xqrs_detect)

[fgr1986]

May I ask why the comparison does not work when the full annotation.sample is passed to the compare_annotations method, and instead annotation.sample[1:] works well?

[cx1111]

Could you give an example of data where this happens?

[fgr1986]

My bad, I pre-filtered the signal in a bad way leading to the error. Please, forget the issue. Sorry for the inconvenience.

[gencee]

great thks

[jbehar]

Hi Chen, @cx1111

Line 114 of the Comparitor class:

self.specificity = float(self.tp) / self.n_ref

I believe this is the sensitivity (TP/(TP+FN)) and not specificity.

Cheers