Quantitative investigation of the effect of measurement sites on deriving respiratory rate from photoplethysmographic pulse waveform

Respiration rate measurement is clinically important for disease diagnosis and therapy monitoring and control. Current clinical techniques of measuring respiration rate are normally using a mask or a tube with a temperature-sensitive sensor to monitor temperatures changes during inhalation and exhalation, but this is not comfortable to wear for long-term monitoring applications. Researchers have attempted to use simple photoplethysmographic (PPG) measurement to derive respiration rate. One unsolved research question is that it is not known where the relatively reliable body site is for measurement of arterial PPG pulse waveforms from which the respiration rate could be derived. This study aimed to investigate the effect of measurement sites on arterial PPG pulse waveform characteristics and the PPG-derived respiration frequency on healthy subjects, and to identify the relatively reliable body site for recording PPG signals and respiration rate measurement. 36 healthy subjects participated in this study. For each subject, one-minute PPG signals were sequentially recorded from each of the six different body sites (finger, wrist under, wrist upper, arm, earlobe, and forehead) under both normal and deep breathing patterns. The recorded PPG waveforms were normalized to derive waveform characteristics including the pulse peak time (Tp) and position, dicrotic notch time (Tn) and position, the reflection index (RI) and the respiratory frequency. The effects of measurement sites and breathing patterns on the waveform characteristics and PPG-derived frequency were finally investigated by analysis of variance with post-hoc multiple comparison. Histograms, regression analysis and Bland-Altman analysis were also performed to investigate the differences between measurement sites. The results showed that the finger was the most reliable site to get analysable PPG signal under both normal and deep breathing conditions, with the percentage of unanalysable recordings less iii than 6% and 14%, while the forehead was the worst site with corresponding 40% and 60% recordings unanalysable. Compared with the measurements from the fingertip, the other five measurement sites produced significantly smaller mean pulse amplitude (all p < 0.001), and longer peak timing (all p < 0.05). The notch point timing derived from the finger PPGs was significantly different with these from the arm (p < 0.001), wrist upper and earlobe (both p < 0.05) under normal breathing, and with that from the earlobe (p < 0.001) under deep breathing. The RI values for the other body sites were significantly different with that from the finger site (all p < 0.001) under both breathing conditions. Regarding the PPG-derived respiratory frequency, under normal breathing, the derived respiratory frequency from the forehead had the smallest bias (-0.011 Hz) with 95% limits of agreement from -0.1 to 0.08 Hz, then from the earlobe, arm, wrist under, finger and wrist upper (-0.05 Hz bias with 95% limits of agreement from -0.24 to 0.14 Hz). Under deep breathing, using the PPGs from the finger and the earlobe derived the most accurate respiratory frequency (100% measurements within error of 0.05 Hz, with a bias of 0.00026 Hz and a 95% limits of agreement from -0.0054 to 0.006 Hz), followed by the forehead and wrist upper, wrist under and finally the arm as the worst position (0.016 Hz bias with a 95% limits of agreement from -0.086 to 0.12 Hz). In summary, this study has provided quantitative evidence that the fingertip was the most reliable body site to measure analysable PPG signal, and that the measurement site had significant effect on PPG waveform characteristics and PPG-derived respiratory rate. This study provided a better understanding of the underlying mechanism of PPG waveform shape measured from different body sites and also a scientific evidence to support the future development of a reliable respiratory monitoring device based on simple PPG measurement.