Computer-assisted diagnosis for chronic heart failure by the analysis of their cardiac reserve and heart sound characteristics

https://doi.org/10.1016/j.cmpb.2015.09.001Get rights and content

Highlights

  • We proposed an innovative LS-SVM based system for diagnosis of CHF.

  • Two novel features of heart sound such as fPSDmax and sub _ EF were proposed.

  • The cardiac reserve indexes and heart sound features were used to diagnose the CHF.

  • The classification performances of LS-SVM, HMM and BP-ANN were compared.

Abstract

An innovative computer-assisted diagnosis system for chronic heart failure (CHF) was proposed in this study, based on cardiac reserve (CR) indexes extraction, heart sound hybrid characteristics extraction and intelligent diagnosis model definition. Firstly, the modified wavelet packet-based denoising method was applied to data pre-processing. Then, the CR indexes such as the ratio of diastolic to systolic duration (D/S) and the amplitude ratio of the first to second heart sound (S1/S2) were extracted. The feature set consisting of the heart sound characteristics such as multifractal spectrum parameters, the frequency corresponding to the maximum peak of the normalized PSD curve (fPSDmax) and adaptive sub-band energy fraction (sub _ EF) were calculated based on multifractal detrended fluctuation analysis (MF-DFA), maximum entropy spectra estimation (MESE) and empirical mode decomposition (EMD). Statistical methods such as t-test and receiver operating characteristic (ROC) curve analysis were performed to analyze the difference of each parameter between the healthy and CHF patients. Finally, least square support vector machine (LS-SVM) was employed for the implementation of intelligent diagnosis. The result indicates the achieved diagnostic accuracy, sensitivity and specificity of the proposed system are 95.39%, 96.59% and 93.75% for the detection of CHF, respectively. The selected cutoff values of the diagnosis features are D/S = 1.59, S1/S2 = 1.31, Δα = 1.34 and fPSDmax = 22.49, determined by ROC curve analysis. This study suggests the proposed methodology could provide a technical clue for the CHF point-of-care system design and be a supplement for CHF diagnosis.

Introduction

Chronic heart failure (CHF) occurs in the situation that heart loses the ability to pump adequate oxygen-rich blood to meet the need of peripheral tissues and organs of the body. This may cause some symptoms such as shortness of breath, tiredness, irregular heartbeats, etc. Compared to the expensive imageological diagnosis and biochemical analysis, it is of great significance to develop a non-invasive, low-cost and convenient detection method for CHF diagnosis.

Many researchers have devoted themselves to the studies on computer-assisted diagnosis for CHF based on the detection and analysis of Electrocardiograph (ECG). Ivanov et al. [1], [2] and Dutta [3] have found there are both a loss of multifractality in heartbeat sequences and ECG of the patients with CHF. The prolongation duration of QRS or wide QRS/T angles could be a predictive indicator of CHF [4], [5]. Skrabal et al. [6] used ECG detection combined with bio-impedance measurement technique to diagnose CHF. However, ECG can only detect the cardiac chronotropic and dromotropic action instead of the cardiac inotropic action that is reduced significantly in CHF patients [7], so it can be seen that single ECG detection for the diagnosis of CHF is insufficient.

Heart sound is very important as it directly reflects the mechanical properties of heart activity [8], [9]. The studies on the relationship between heart sound and cardiac contractility indicate that the amplitude of the first heart sound (S1) is positively correlated with the maximum rise rate of left ventricular pressure (r = 0.9551, p < 0.001) and the amplitude of S1 is also closely related to the strength of cardiac contractility [10], [11]. This has suggested that the amplitude of S1 can reflect the level of cardiac contractility. The most important aspect of cardiac dysfunction in heart failure is not the depressed cardiac performance observed at basal resting state but rather the loss of cardiac reserve (CR) [12], [13], which is manifested in the decrease of cardiac contractility, so the detection and analysis of heart sound and the measurement of CR could provide important clues for the diagnosis of CHF. Based on the relationship between heart sound and cardiac contractility, an noninvasive and quantitative method for the assessment of CR has been proposed by our group [14], [15]. Some diagnostic techniques such as real-time transmission of the phonocardiogram (PCG) through the Internet and computer-assisted auscultation were developed [16], [17]. The application of CR indexes in monitoring and evaluating heart function for gestational woman was implemented [18]. However, until now, the studies about the application of CR in the diagnosis of CHF have not been reported, and the utilizations of heart sound characteristics for the diagnosis of CHF are few, except that an appearance of the third heart sound is regard as a highly specific and none sensitive marker for the diagnosis of CHF [19], [20].

In this paper, an intelligent diagnosis system for CHF diagnosis was proposed, the schematic of which is shown in Fig. 1. It consists of acquisition system (hardware) and decision support system (software). The acquisition system includes sensor, acquisition circuit and computer device shown in Fig. 2. The decision support system is embedded in the computer. This paper emphatically introduces the decision support system that includes the following parts. The preprocessing is implemented based on amplitude normalization and modified wavelet packet denoising methods. The CR indexes such as the ratio of diastolic to systolic duration (D/S) and the amplitude ratio of the first to second heart sound (S1/S2) combined with three heart sound characteristics such as the frequency corresponding to the maximum peak of the normalized PSD curve (fPSDmax), adaptive sub-band energy fraction (sub _ EF) and multifractal spectrum parameter were proposed to structure a diagnostic feature set. The self-developed cardiac reserve monitor software (CRM version1.0, Chongqing University and Bo-Jing Medical Informatics Institute, China) was used to measure the CR indexes, and the heart sound characteristics were extracted based on maximum entropy spectra estimation (MESE), empirical mode decomposition (EMD) and multifractal detrended fluctuation analysis (MF-DFA) methods which are good at the analysis of non-stationary and non-linear physiological signal [21], [22], [23]. The LS-SVM was determined as the classifier of proposed system by the comparison of performances with back-propagation artificial neural network (BP-ANN) and hidden markov model (HMM). A dataset collected from the healthy volunteers and CHF patients was used to verify the proposed system. In addition, statistical analysis methods such as t-test and ROC curve were conducted to suggest the diagnosis thresholds. The purpose of our study is to explore a new effective computer-assisted diagnosis technique for the diagnosis of CHF.

The outline of this paper is organized as follows. Section 2 describes the detail of proposed diagnostic system including the methodologies of preprocessing, feature extraction and identification. Section 3 represents the statistical result of CR indexes and heart sound characteristics and the comparison of diagnostic performance among LS-SVM, BP-ANN and HMM. Section 4 discusses the differences of diagnostic indexes and characteristics between the healthy and CHF patients as well as the advantage and limitation of the proposed system. Section 5 gives the conclusion and future work of the study. The methodology framework of this paper is shown in Fig. 3.

Section snippets

Study participants and clinic trial description

The subjects consist of 88 healthy volunteers (college students and teachers) as controls and 64 CHF patients, who knew and signed the informed consent forms. The patients with CHF include the patients with heart failure with reduced ejection fraction (HFrEF) and heart failure with preserved ejection fraction (HFpEF), which are confirmed by the experienced cardiologists. The CHF patients with left ventricular ejection fraction (LVEF) more than 50% are considered as the patients with HFpEF, and

The analysis result of CR indexes and heart sound characteristics

LVEF is a worldwide acknowledged index to reflect the differences of heart pump function between the healthy and people with CHF [37]. It was just used to verify the diagnosis of CHF in this study. CR indexes extracted from the healthy volunteers and CHF patients are listed in Table 3. The D/S and S1/S2 values of control group are higher than those of CHF group, while the HR values of the control group are lower.

Table 4 presents the calculated results of multifractal spectrum parameters. The

The differences of CR indexes between the control and CHF group

Cardiac reserve is regarded as an important physiological function base of the fitness and exercise performance of human beings [38]. This paper has investigated the differences of CR indexes between the healthy and CHF patients. D/S value is the time index of heart perfusion reserve, and it emphasizes the significance of the appropriate ratio of diastole to systole duration to maintain the healthy function of heart pump [15]. Because the most part of coronary blood flow occurs during the

Conclusion

In this paper, we proposed an intelligent system for the diagnosis of CHF based on LS-SVM. The CR indexes such as D/S and S1/S2 and heart sound characteristics such as Δα, fPSDmax and adaptive sub _ EF are constituted diagnosis features as the input of classifier. A dataset collected from the healthy volunteers and CHF patients was used to verify the proposed system. The LS-SVM classifier with satisfactory accuracy, sensitivity and specificity was selected through the comparison with BP-ANN and

Conflict of interest statement

None.

Acknowledgement

This study is supported by the National Natural Science Foundation of China (No. 30770551).

References (51)

  • Y. Zheng et al.

    A novel hybrid energy fraction and entropy-based approach for systolic heart murmurs identification

    Expert Syst. Appl.

    (2015)
  • J.W. Kantelhardt et al.

    Multifractal detrended fluctuation analysis of nonstationary time series

    Phys. A: Stat. Mech. Appl.

    (2002)
  • D. Tsujinishi et al.

    Fuzzy least squares support vector machines for multiclass problems

    Neural Netw.

    (2003)
  • M.A. Hanson

    On sufficiency of the Kuhn–Tucker conditions

    J. Math. Anal. Appl.

    (1981)
  • R.S. Vasan et al.

    Congestive heart failure in subjects with normal versus reduced left ventricular ejection fraction: prevalence and mortality in a population-based cohort

    J. Am. Coll. Cardiol.

    (1999)
  • B.P. Hsieh et al.

    The amplitude ratio of the first to second heart sound is reduced in left ventricular systolic dysfunction

    Int. J. Cardiol.

    (2010)
  • J. Dean et al.

    Arrhythmia in heart failure: role of mechanically induced changes in electrophysiology

    Lancet

    (1989)
  • S. Sun et al.

    Segmentation-based heart sound feature extraction combined with classifier models for a VSD diagnosis system

    Expert Syst. Appl.

    (2014)
  • S. Choi et al.

    A novel cardiac spectral segmentation based on a multi-Gaussian fitting method for regurgitation murmur identification

    Signal Process.

    (2014)
  • S. Ari et al.

    Detection of cardiac abnormality from PCG signal using LMS based least square SVM classifier

    Expert Syst. Appl.

    (2010)
  • E. Avci

    A new intelligent diagnosis system for the heart valve diseases by using genetic-SVM classifier

    Expert Syst. Appl.

    (2009)
  • R. Moraes et al.

    Document-level sentiment classification: an empirical comparison between SVM and ANN

    Expert Syst. Appl.

    (2013)
  • P.C. Ivanov et al.

    Multifractality in human heartbeat dynamics

    Nature

    (1999)
  • P.C. Ivanov et al.

    From 1/f noise to multifractal cascades in heartbeat dynamics

    Chaos

    (2001)
  • S. Dutta

    Multifractal properties of ECG patterns of patients suffering from congestive heart failure

    J. Stat. Mech.: Theory Exp.

    (2010)
  • Cited by (65)

    • Recent Trends in Computer-Aided Diagnostic Systems for Skin Diseases: Theory, Implementation, and Analysis

      2021, Recent Trends in Computer-aided Diagnostic Systems for Skin Diseases: Theory, Implementation, and Analysis
    View all citing articles on Scopus
    View full text