APPLYING WEIGHTED PULSE DECOMPOSITION ANALYSIS WITH SIGNAL QUALITY INDEX TO FINGER PHOTOPLETHYSMOGRAM FOR ESTIMATION OF BLOOD PRESSURES. (April 2021)

• Record Type: Journal Article
• Authors: Wang, Tzung-Dau; Guo, Jia-Wei; Tsai, Pei-Yun; Lin, Hung-Ju; Wu, An-Yeu
• Is Part Of: Journal of hypertension. Volume 39(2021)e-Supplement 1
• Publication Date: 2021-04
• Subjects: Hypertension -- Periodicals; Hypertension -- Periodicals; 616.132005
• DOI: 10.1097/01.hjh.0000745096.58648.0e
• Languages: English
• ISSNs: 1473-5598

Abstract

Abstract : Objective: Pulse decomposition analysis of photoplethysmography (PPG) signals can be used to estimate blood pressures. This study aimed to generate stable component waves by introducing weighting to the informative PPG portion and to ensure the reliability of the decomposed outputs assessed by signal quality index (SQI) and its relations to blood pressures. Design and method: We adopted weighted least squares criterion for the pulse decomposition analysis. Five Gaussian component waves were used. The systolic wave and diastolic wave were obtained by summing up the first three component waves and the last two component waves, respectively. The stiffness index (SI) was calculated by the time difference of the systolic wave peak (t_S) and the diastolic wave peak (t_D). The mean square error (MSE) between the synthesized PPG pulse and the normalized PPG pulse was adopted as SQI to evaluate decomposed results. The effect of weighting was investigated and compared by averaged variance of t_S and t_D. Proper SQI was examined to strike a good balance between the pass rate and the degradation of correlation coefficients of blood pressures to SI. Results: PPG data were collected from 20 subjects, with systolic blood pressures (SBP) and diastolic blood pressures (DBP) from 99 to 158 mmHg and 60 to 110 mmHg, respectively. A total of 2442 PPG pulses were analyzed. The averaged variances of t_S and t_D were reduced from 9.69 to 6.1 ms and 15.31 to 9.89 ms if weighting was adjusted from 1 to 80. The MSE for SQI was set around 2 × 10(-3), which could achieve pass rate of 98.7%. The degradation of correlation coefficients of body weight divided by SI, a parameter reflective of vascular stiffness, to SBP and DBP was less than 0.02, and was greater than that of body weight divided by SI obtained by conventional Gaussian decomposition (r for SBP, 0.51 vs 0.44; r for DBP, 0.49 vs 0.40). Conclusions: With the introduction of weighting, the decomposed PPG component waves became stabilized. SQI in terms of MSE for weighted pulse decomposition analysis ensured reliable outputs and enhancing its relations to blood pressures.