Estimation of baroreflex sensitivity by Gaussian average filtering decomposition. (July 2021)

• Record Type: Journal Article
• Authors: Lin, Yue-Der; Zida, Serge Ismael
• Is Part Of: Biomedical signal processing and control. Volume 68(2021)
• Publication Date: 2021-07
• Subjects: α index -- Autoregressive (AR) -- Baroreflex sensitivity (BRS) -- Gaussian average filtering decomposition (GAFD) -- Power spectral density (PSD) -- Welch's periodogram; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2021.102576
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: A dedicated algorithm based on Gaussian moving-average filter is proposed to estimate baroreflex sensitivity (BRS). The proposed algorithm is computationally efficient. The BRS values derived by the proposed algorithm have no statistical difference compared with those derived by spectral methods. The proposed algorithm is potential for the real-time estimation of BRS in clinical settings. Abstract: Baroreflex sensitivity (BRS) has been verified to be an important index in clinical settings. This article proposes a dedicated algorithm termed as Gaussian average filtering decomposition (GAFD) to estimate BRS in an efficient way. The proposed algorithm is similar to wavelet decomposition except the Gaussian averaging filter of predetermined cutoff frequencies and the variable window length based on the number of extremes in the signals. Only moving average and squared summation are required, which makes the proposed method efficient in computation. In addition, the proposed technique is performed thoroughly in time domain but can acquire the information similar to that derived by conventional spectral methods. The EuroBavar dataset is adopted in all computer experiments to evaluate the performance of the proposed algorithm. In addition to the proposed approach, the BRS values are also estimated by two conventional spectral methods, Welch's periodogram and autoregressive (AR) power spectral density (PSD), for comparison. The BRS values derived by the proposed method have compared with those derived by spectral methods and have been verified to be of no statistical difference using Wilcoxon rank sum test and of excellent agreement using intraclass correlation coefficient (ICC). The proposed algorithm is computationally efficient and is feasible to be incorporated into the embedded medical devices, which makes the proposed method potential for the real-time estimation of BRS in clinical settings.