Empirical mode decomposition for respiratory and heart rate estimation from the photoplethysmogram

We introduce a method based on empirical mode decomposition (EMD) to estimate both respiratory rate (RR) and heart rate (HR) from the photoplethysmographic (PPG) signal obtained from pulse oximetry. The spectral analysis of the EMD applied to the PPG signal was used to extract two signals, the respiratory and cardiac modulations respectively. On these modulated signals, an additional spectral analysis was applied to calculate their frequency peaks. To improve spectral resolution a parametric power spectral analysis based on autoregressive modelling was selected. The frequency peak found in the respiratory and cardiac signals reflects RR and HR, respectively. The PPG signals were analysed using a 1-min sliding window with 50% overlap. The RR and HR estimation accuracy was assessed using the unnormalized root mean square (RMS) error. Median errors (quartiles) were calculated to account for the non-normal RMS distribution. The test dataset consisted of 8-min PPG and capnometric signals from 29 paediatric and 13 adults cases (42 subjects in total) containing reliable recordings of either spontaneous or controlled breathing. A research assistant manually labelled the signals. The reference RR (from capnogram) and HR (from PPG) were manually extracted. The median RMS error (quartiles) obtained for RR was 3.5 (1.1, 11) breaths/min and for HR was 0.35 (0.2, 0.59) beats/min. Therefore, the spectral analysis of the respiratory and cardiac signals extracted through EMD, introduces a useful method to estimate and monitor RR and HR simultaneously from the PPG signal obtained from pulse oximetry.