Biomedical Signal Processing and Control 46 (2018) 260–267 Contents lists available at ScienceDirect Biomedical Signal Processing and Control journal homepage: www.elsevier.com/locate/bspc Use of cardiorespiratory coherence to separate spectral bands of the heart rate variability Mohamed Daoud a,b , Philippe Ravier a,∗,1 , Olivier Buttelli a a Univ. Orléans, INSA-CVL, Laboratoire PRISME, EA 4229, F45072, Orléans, France b Université de Mostaganem, Laboratoire Signaux et Systèmes, Algeria a r t i c l e i n f o Article history: Received 22 October 2017 Received in revised form 14 June 2018 Accepted 1 August 2018 Keywords: HRV Respiration Cardiorespiratory interaction HRV spectral boundaries Time-frequency coherence function Instantaneous center frequency a b s t r a c t Regulation of heart rate variability (HRV) is dependent on the autonomic nervous system (ANS). The regulation is achieved by the two ANS activity branches distribution, namely the parasympathetic and the sympathetic activities. Their fluctuations can be observed in the HRV signal spectral domain through respective dissociated frequency domains. Power spectrum parameters in low frequency (LF) and high frequency (HF) domains are classically estimated using predetermined fixed frequency ranges which are [0.04–0.15 Hz] band for LF domain and [0.15–0.4 Hz] band for HF domain. However, the 0.15 Hz frequency threshold is controversial in dynamic situations or positions changing. Several studies advanced that the observed cardiac and respiratory oscillation interactions could reflect the functionality of the ANS. These interactions can be estimated by the cardiorespiratory spectral coher- ence tool which quantifies the existence and strength of linearity between the two modes. We propose in this paper to jointly use the heart rate variability (HRV) and respiration modes by exploiting the car- diorespiratory spectral coherence tool for the time-varying frequency threshold definition between LF and HF bands. The instantaneous center frequency computed on the time-frequency coherence estimator is the proposed variable threshold (VT). The VT method was implemented on both simulated data and real data acquired from the experimental orthostatic test protocol with 10 healthy subjects. The classical 0.15 Hz fixed threshold (FT) and the individual time-dependent spectral boundaries method (AT) using the lower limit of the respiratory HF band as a time-varying threshold are used for comparison. The proposed VT strategy makes it possible to obtain: (i) a strengthened data-dependent definition of boundaries for the HF domain; (ii) instantaneous estimations of the LF and HF band energies; (iii) a better discrimination between the two positions SUPINE and UPRIGHT of the classical orthostatic test. © 2018 Elsevier Ltd. All rights reserved. 1. Introduction The heart rate variability (HRV) analysis constitutes a power- ful tool to help the understanding of the regulatory mechanisms of heart by the autonomic nervous system (ANS) [1]. The HRV is usu- ally calculated by analyzing the oscillations of the peak-to-peak RR intervals sequence derived from the electrocardiogram, R being the time peak location of each beat. The HRV is partly due to the heart rate control mode, which is influenced by the vegetative nervous system. This variation is controlled by two branches of the ANS: the parasympathetic (vagal nerve) and the sympathetic branches. The effects of both ANS activities can be observed in respective disso- ∗ Corresponding author. Tel.: +33 238494863. E-mail address: philippe.ravier@univ-orleans.fr (P. Ravier). 1 www.univ-orleans.fr/prisme. ciated frequency domains: the parasympathetic branch in the high frequency domain (HF) and both branches in the low frequency (LF) domain [2]. In HRV analysis, LF and HF power spectrum parame- ters are classically estimated using predetermined fixed frequency ranges which are [0.04 - 0.15 Hz] band for LF domain and [0.15 - 0.4 Hz] band for HF domain. However, other factors than only the vegetative activity can interact with HRV. Notably, several studies have reported an influence of breath- ing on this heart rate, since the first works of Hirsch and Bishop [3]. The respiratory oscillations may influence the content of these fre- quency bands [4]. Mainly HF band was influenced by this activity [5] and consequently it was reported a strong coherence between the respiratory rhythm and the HF domain [6]. It is the reason why was highlighted the necessity of properly considering the respiratory signal as complementary information for any ANS study when using HRV analysis. Moreover, another issue in the LF and HF domain def- https://doi.org/10.1016/j.bspc.2018.08.003 1746-8094/© 2018 Elsevier Ltd. All rights reserved.