Fig. 3: Characterization of random rhythmic and arrhythmic templates for mechanical beating recognition under 80 nM astemizole treatment in the short term.

a Extraction and reconstitution of one rhythmic template and two arrhythmic templates. Templates were randomly extracted from typical mechanical beating signals by valley detection and were then reconstituted to match the data length of the target signals by spline interpolation and resampling. b–c Matching of the templates with the typical signals in (a) by correlation analysis. Templates and signals with similar profiles showed a high correlation (>0.98) in the short term, while templates and signals with different profiles showed a low correlation (<0.94). The correlation (<0.9) between the rhythmic template and arrhythmic signals was lower than that between the arrhythmic template and arrhythmic signals (>0.9). n ≥ 21 from three data segments; ****p < 0.0001. e–g Universality test of the random mechanical beating templates by analysis of the correlation between similar signals in the short term. The correlation between the rhythmic template and rhythmic signals was over 0.995. The correlation between arrhythmic template I with two positive peaks and arrhythmic signals I was over 0.992. The correlation between the arrhythmic template II with three positive peaks and arrhythmic signals II was over 0.982