Extraction of photoplethysmographic waveform variability by lowpass filtering

Cardiovascular variability is known to provide useful physiological information about autonomic function and peripheral vascular tone. Previous studies have used systolic values (peaks) or diastolic values (troughs) in the photoplethysmographic signal (PPG) to represent the variability in the finger pulse waveform. However, the feature detection process is error prone and often requires manual correction which is time consuming. The current study has proposed the lowpass filtering method as an alternative means to extract the variability signal. The similarities between the lowpass filtered spectrum and the spectra produced by other representation methods were assessed quantitatively via the computation of normalized cross-correlations. Results showed that the lowpass filtered signal produced a variability spectrum which was nearly identical to that of the pulse waveform mean value (correlation = 0.996), and was highly correlated with the trough and the peak variability spectra (correlation > 0.9). Compared with feature extraction methods, the lowpass filtering method is much simpler and computationally efficient to implement. In addition, the lowpass filtering method can be applied in conjunction with signal decomposition techniques such as principal component analysis (PCA) to better quantify sympathetic change.