Analyzing seismocardiographic approach for heart rate variability measurement

https://doi.org/10.1016/j.bspc.2021.102793Get rights and content

Highlights

  • For AO peak detection, the performance of the previously proposed MVMD technique is improved by adding a decision-rule-based post-processing scheme.

  • Temporal and spectral HRV parameters are estimated from the tachograms created from consecutive AO–AO intervals.

  • The performance of the proposed method is tested and validated with healthy subjects at different postures and physiological conditions.

  • The performance validation is done with the heart cycles and HRV parameters derived from reference ECG signals. Various statistical measures are used for the validation purposes, including normalized cross correlation (NCC), absolute error, mean error, root mean square error (RMSE), and limits of agreement (LOA).

Abstract

As a vital risk stratification tool, heart rate variability (HRV) has the ability to provide early warning signs for many life-threatening diseases. This paper presents a study on reliable cardiac cycle extraction and HRV measurement with a seismocardiographic (SCG) method. Like R-peaks in an ECG, the proposed method relies on peaks corresponding to aortic valve opening (AO) instants in an SCG signal. Due to better reliability and accessibility, the SCG signal is selected for the study. Initially, the prominent AO peaks in an SCG signal are estimated using our previously proposed modified variational mode decomposition (MVMD) based approach. In the present method, the detection performance of AO peaks is improved by employing a decision-rule-based post-processing scheme. Subsequently, tachogram of AO–AO intervals is used for the estimation of HRV parameters. A set of real-time signals collected in various physiological conditions and the SCG signals taken from a publicly available standard database are used to test and validate the proposed method. Experimental results clearly tell that the cardiac intervals obtained from the SCG signal using the proposed method can be used for HRV analysis. Also, the resulted parameters of HRV analysis on ECG and SCG exhibit strong correlation and agreement that shows the effectiveness of the proposed method.

Introduction

Heart rate variability (HRV) analysis is usually employed to assess the autonomic nervous system (ANS) functioning in cardiac studies. The HRV is a physiological phenomenon that describes the oscillation in consecutive heart cycles [1]. Both sympathetic and parasympathetic activities can be apparently characterized by this tool. Clinically, HRV has proven its abilities in major applications, such as diagnosis of sudden cardiac death due to acute myocardial infarction and early prediction of diabetic neuropathy [1], [2]. Besides these, HRV has been investigated for several cardiac diseases, physical exercise, renal failure, stress conditions, sleep disorder, age, gender, drugs, alcohol consumption, and smoking [2], [3], [4]. The HRV analysis gives many temporal, spectral, and nonlinear indices solely relying on beat-to-beat interval traces.

The prevalence of HRV estimation has led to increased development of several commercial and non-business software tools [2], [5]. ECG is considered as a standard tool for HRV estimation. But it requires the attachment of multiple electrodes to the skin surface with an adhesive gel for more electrical conduction. Thus, a user-discomfort, experienced due to this, becomes an obstacle for long-term continuous monitoring. Additionally, its recording requires an expert for electrode placement and clinical interpretation, and therefore, it is difficult to use effectively outside the clinical setup. There has been a continuous improvement in the sensing devices used for cardiac signal acquisition to overcome these limitations. Seismocardiography (SCG) is immensely used in recent years due to technological advancements and miniaturization of accelerometers [6], [7]. In contrast to other cardiac signals, an SCG is able to capture more diagnostic information and has been studied for various clinical and non-clinical applications [8]. Usually, a fiducial point corresponding to aortic valve opening (AO) creates a prominent peak in an SCG cycle (refer Fig. 4), and hence, it has been studied more for heartbeat extraction and HRV applications. For a reliable HRV analysis, an accurate estimation of AO peaks is quite essential. Along with AO peaks, other SCG points such as isovolumetric contraction (IVC) has also been investigated by the researchers for HRV analysis [9], [10], [11], [12], [13], [14], [15]. Ramos-Castro et al. performed HRV analysis for the first time ever on SCG signals [9]. Ramos-Castro et al. [9] and Landreani et al. [12] showed that IVC location of an SCG signal, acquired through a smart-phone, can be used for this purpose. For HRV estimation, the AO peak locations in the SCG signal have been investigated quite well [10], [11], [14]; nevertheless, they require to locate R-peaks of a concurrent ECG signal for AO peak detection. Usually, they employ the ECG-SCG windowing technique to estimate their locations. Recently, different cardiac signals, including SCG, PCG, PPG, and piezoplethysmocardiogram, are compared with their concurrent ECG signal for HRV analysis [15]. The same peak detection logic is applied for all of them to extract the heart cycles. The method uses a simple thresholding based peak detection scheme using a Matlab function Findpeaks with a refractory period of 500 ms. However, it is not a suitable technique for signals having a large extent of noises, interferences, and interbeat and intrabeat variabilities. Hence, most of the existing HRV analysis methods based on SCG signals do not employ proper and reliable fiducial point estimation techniques. Additionally, they do not use standalone methods for this purpose. As a result of this, not only user-comfortability decreases, but also overall cost increases.

To address these issues, we proposed an SCG-based algorithm for HRV analysis, which entirely depends on the AO peaks of an SCG signal. The present approach uses our previously proposed AO peak detection method [16], which has shown explicitly good performance on public and private databases. Also, it does not rely on any other cardiac signal as a reference. However, the AO peak detection performance may be improved by applying an appropriate post-processing scheme. The main objective is to validate our AO detection method for the HRV measurement. The method performs time-domain and frequency-domain HRV analysis, and the results derived from SCG are cross-validated with ECG as a reference. The performance results are validated using absolute error and normalized linear cross-correlation measures. The paper is organized as follows: Section 2 presents the description of databases used and the proposed method for the estimation of HRV parameters. Experimental results are discussed in Section 3. Finally, conclusions are drawn in Section 4.

Section snippets

Materials and methods

In this study, a publically available CEBS database (at PhysioNet archive) [17], [18], [19] and a private database are used for performance analysis. The CEBS database consists of 20 multi-channel recordings acquired from twenty healthy subjects. Each recording possesses four channels: ECG (I and II), respiration signal, and SCG. Recordings were taken in still awake condition and supine position of subjects in three subsequent states: before listening to music at basal state for 5 min, music

Results and discussion

In the proposed method, the AO peaks of an SCG are detected initially along with the identification of R-peaks in the concurrent ECG signal. An experimental result for the detection of these peaks is shown in Fig. 4(a) and (b). The derived AO–AO and RR intervals are used to create traces of cardiac cycles. Before going deep further into HRV analysis using AO–AO interval time-series, it is interesting to perform the reliability-test with RR intervals. For instance, it is observed that the

Conclusion

In this work, we have proposed a reliable SCG-based HRV estimation method. For this purpose, our MVMD-based AO instant detection method integrated with the proposed post-processing scheme was investigated. All the experiments were performed on a publicly available CEBS database and our own registrations acquired at various physiological variations. The method has been validated for a robust heartbeat extraction by comparing the AO–AO intervals with RR intervals extracted from a reference ECG.

Authors’ contribution

Tilendra Choudhary: conceptualization, methodology, reviewing and editing. Mousumi Das: writing-original draft preparation, editing. L.N. Sharma: supervision, reviewing and editing. M.K. Bhuyan: supervision.

Acknowledgments

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. The authors would like to thank the anonymous reviewers in advance for their useful suggestions and comments.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References (25)

  • M.J. Tadi et al.

    Seismocardiography: toward heart rate variability (HRV) estimation

    Medical Measurements and Applications (MeMeA), IEEE International Symposium

    (2015)
  • F. Landreani et al.

    Beat-to-beat heart rate detection by smartphone’s accelerometers: validation with ECG

    38th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC)

    (2016)
  • Cited by (0)

    View full text