Figure 1.
Description of the Bivariate Phase Rectified Signal Averaging (BPRSA) method.
Panel a) Uterine Activity by EHG (trigger signal): each data point is the average of raw signal every 2 seconds. Each black dot represents an anchor point (EHG increases). Panel b) fetal heart rate by ECG (target signal). The empty circles represent anchor points derived from the EHG signal. Panel c) Segments of fetal heart rate to average, centered around the anchor point (time windows of 200 seconds, each data point(i) is the average of fHR every 2 seconds). Panel d) phase rectification at the anchor points and signal averaging. Panel e) BPRSA graph representing PRSA transformation of Uterine Activity (grey line) and BPRSA transformation of fetal heart rate (bold line).
Figure 2.
Example of bivariate phase rectified signal averaging (BPRSA) graphs:
grey line: univariate PRSA transformation of electrohysterogram increases (EHG↑) time series; black line: bivariate PRSA transformation of fetal heart rate (fHR) values in correspondence to EHG↑. The time axis, centered around time 0 (aligned anchors), allows to observe the oscillation of the signals before and after the triggering event (time windows of 200 seconds, each data point(i) is the average of PRSA-signals every 2 seconds). Panel a: presence of coupling (black line oscillation); panel b: absence of coupling (flat black line).
Figure 3.
Detection of cross-correlations by cross power spectral density (CPSD) analysis applied to PRSA transformed signals: a
) 30-minutes electrohysterogram and fetal heart rate (EHG-fHR) recording; b) BPRSA graph denoting presence of coupled periodicities between uterine PRSA (grey line) and fHR BPRSA (black line); c) CPSD analysis of PRSA transformed signals. The coefficient of coherence CPRSA shows a significant correlation at the UC frequency domain (CPRSA = 0.81).
Figure 4.
Comparison of BPRSA with conventional cross spectral analysis: example of detection of coupling by BPRSA.
a) 30 minutes recording of electrohysterogram (EHG, grey line) and fetal heart rate (fHR, black line) free of signal loss; b) cross power spectral density (CPSD) analysis of original signals, CRAW<0.5 at the uterine contraction (UC) frequency domain; c) BPRSA graph showing coupled periodicities; d) CPSD analysis of PRSA signals, the CPRSA>0.5 at the UC frequency domain proved a significant correlation.
Figure 5.
Comparison of BPRSA with conventional cross spectral analysis; example of absence of coupling by BPRSA.
a) 30 minutes recording of electrohysterogram (EHG, grey line) and fetal heart rate (fHR, black line) free of signal loss; b) cross power spectral density (CPSD) analysis of original signals, CRAW<0.5 at the uterine contraction (UC) frequency domain; c) BPRSA graph displaying absence of coupled periodicities (flat BPRSA signal, black line); d) CPSD analysis of PRSA signals, the CPRSA<0.5 at the UC frequency domain proved the absence of significant correlation.
Figure 6.
Comparison between coefficients of coherence
provided by conventional cross-spectral analysis (CRAW, median 0.29, interquartile range 0.17–0.49) and by bivariate Phase-Rectified Signal Averaging (CPRSA median 0.79, interquartile range 0.69–0.91); p<0.0001.