A peak detection method for identifying phase in physiological signals. (January 2017)

• Record Type: Journal Article
• Authors: Mitrou, Nicholas; Laurin, Alexandre; Dick, Taylor; Inskip, Jessica
• Is Part Of: Biomedical signal processing and control. Volume 31(2017)
• Page Start: 452
• Page End: 462
• Publication Date: 2017-01
• Subjects: Phase estimation -- Wavelet transform -- Peak detection -- Hilbert transform -- Cardiac baroreflex -- Renal autoregulation; 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.2016.07.001
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: A simplified algorithm for detection of phase in physiological signals is presented. The peak detection algorithm performed as well as current gold-standard techniques in terms of accuracy. Peak detection has a faster computation time than wavelet transform. Abstract: Introduction: To understand the integrated behavior of biological systems, the interactions between their constituent parts are often studied. For example, the interaction between blood pressure and heart rate reveals information about the cardiac baroreflex. For the purpose of characterizing relationships between physiological signals, it is useful to identify phase either as a primary outcome or as an intermediate step to obtain other relevant secondary indices. Existing methods for phase estimation in physiological signals often suffer from a lack of thorough description and standardization, which renders reproducibility and interpretation difficult. A relatively simpler peak detection algorithm was compared to the gold standards of wavelet and Hilbert transforms for its ability to obtain phase. Methods: The accuracy and computation time of the peak detection algorithm was compared to the gold standard methods in silico by applying all three to data of known phase, and signal-to-noise ratios from −20 to 5 dB. We then compared the performance of the peak detection method to the Hilbert and wavelet methods by applying each to four different types of in vivo data. Results: The peak detection technique is less susceptible to noise and over 10 times faster, computationally, than the wavelet technique. Application to in vivo physiological data shows that equivalent results are obtained from each technique. Conclusions: The peak detection method can be used to obtain phase in physiological signals, provide a clearer and more direct interpretation, and be more easily reproducible. Because of its design features, peak detection could also be used to identify individual oscillations in relevant signals, as well as to obtain amplitudes and direct time delays.