Near-Real-Time Detection of Pulse Oximeter PPG Peaks Using Wavelet Decomposition ?

摘要

Pulse oximeters are frequently used to provide real time measurements of heart rate and blood oxygen saturation(SpO2). SpO2is calculated by taking the ratio of the AC to DC components of the photoplethysmograph (PPG) signal measured by the pulse oximeter. For accurate estimation ofSpO2,the AC component needs to be extracted from the signal through signal processing, where accurate peak detection is a crucial, difficult element. This paper investigates the use of the wavelet transform for real time signal processing to detect peaks that could be unintentionally attenuated through more conventional filtering methods. Four mother wavelets (Daubechies 3, symlets 2, coiflets 3 and reverse biorthogonal 1.5) were tested against each other to determine the wavelet with the best representation of the PPG signal in a noisy environment (SNR of 6.44). The reverse biorthogonal (rbio1.5) mother wavelet was found to better represent the PPG signal with a specificity of 0.97 and a sensitivity of 0.97. Further research into the decomposition depth of the rbio1.5 wavelet resulted in an optimal depth of 3, with the2ndand3rdlevels being used for reconstruction of the signal. Using a wavelet length of 128 samples resulted in a time delay of 2.56 seconds. This time delay is well within clinical requirements for near real-time-signal analysis involving these devices.