Single channel high noise level ECG deconvolution using optimized blind adaptive filtering and fixed-point convolution kernel compensation

https://doi.org/10.1016/j.bspc.2019.101673Get rights and content

Abstract

An electrocardiogram (ECG) is used to record the electrical activity of the heart. However, ECG signals are susceptible to the noise from various sources which increases the probability of misinterpretation and can affect the diagnostic process. Traditional noise cancellation techniques, which uses finite and deterministic coefficient, are not efficient, since the ECG signals are non-stationary. Thus, adaptive filters are commonly utilized on such signal as they can adjust their coefficient according to the changing nature of non-stationary signal. Adaptive algorithms still have a disadvantage that they require the model of noise or desired signal. In this paper, a novel algorithm is introduced based on fixed-point convolution kernel compensation for finding a model for using an adaptive filter; then a recursive least square method is used for completing steps of deconvolution of the ECG signal. The deconvolution method can be used for denoising ECG signals in very low signal to noise ratio circumstances and also can be used in blind source separation applications such as separation of fetal ECG from maternal ECG. ECG signals were utilized in this study are taken from the MIT-BIH Arrhythmia database for showing the performance of the algorithm on denoising applications. The results demonstrate that the proposed algorithm renders a much-improved performance in removing the noise from ECG signals, especially in a scenario where signal to noise ratio is negative. Moreover, the noninvasive fetal ECG dataset (NI-FECG) provided by Physionet is also used for fetal ECG extraction by a single thoracic channel. By comparing fetal ECG extraction methods in the literature and the proposed method, it reveals that the proposed method can extract the QRS complex of fetal ECG by a single thoracic channel as accurate as other methods which use abdominal channels.

Introduction

The electrocardiogram (ECG) is an essential tool used for diagnosis of heart disorders. It is a non-invasive recording of heart activity that is typically measured by connecting electrodes on the surface of the specific parts of the human body (chest, arms, hands or legs) [1]. However, with the introduction of new technology (Body Area Network or Body Sensor Network), wearable or implantable devices are getting popular which can monitor the vital signs continuously [2]. Although the introduction of these wearable devices have revolutionized personal healthcare, the obtained ECG signal from these devices is significantly distorted with noises (baseline drift and motion artifacts) [2]. Since ECG can aid in the comprehensive analysis of the heart activity of a patient and can also help in detecting aberrations such as cardiac infractions, unequal beat intervals, the obtained signal should be very accurate. The signals obtained at the surface of the body is weak, have low amplitude and is often susceptible to noise from different sources. These noises are typical results of basic measurement and instrumentation faults. The noises typically arise from power line interference, white noise, electrode contact noise, muscle contraction, baseline wandering, and instrument noise, motion artifacts, electrosurgical noise [3]. Even a slight variation in the obtained ECG waveform can impair the understanding of the heart condition of the patient and can affect the treatment procedure [4,5]. Thus, it is necessary to remove noise as it will help in obtaining the data which is an accurate representation of the heart’s functionality.

Different types of digital filters such as Finite Impulse Response and Infinite Impulse Response filter [[5], [6], [7]] have been utilized in order to eliminate ECG noises. These filters work when parameters of signals and channels are well-known. However, they cannot remove all type of noises efficiently, when signals are nonstationary [8]. In this regard, adaptive filters are introduced to adapt coefficients of filter according to signal changes in time. Nevertheless, adaptive filters have some disadvantages, e.g., the convergence rate is highly depended on the power spectral density of the input signal [9]. Hence, if the power spectrum of signals has a flat and uniform component in all available frequencies which means that the input signal filter is white, the convergence rate of minimum mean-square is excellent. However, colored noise, will drop the efficiency substantially. Therefore, the least mean squares (LMS) and Recursive least square (RLS) which converge to Weiner optimal solution are designed to handle narrowband frequency, and according to the literature [10], even a primary adaptive filter such as LMS can reduce the noise from nonstationary signals. Another drawback of an adaptive filter is the prior knowledge about the desired signal or noise which is paramount [10]. Occasionally, using a template of a well-known shape of ECG epoch as the desired signal, can reduce the chance of detecting abnormality and artifacts [11]. Therefore, blind source separation methods such as independent component analysis (ICA) and principal component analysis are emerged for ECG noise removal; However, in meager SNR circumstances, these methods are not sufficient and need further post-processing [[12], [13], [14], [15]].

In this paper, a robust algorithm is introduced for deconvolution of ECG signal. The proposed algorithm is tested in two ways. First, the proposed algorithm is utilized for single-channel ECG denoising in very low signal to noise ratio condition. Second, the proposed method is used for extracting QRS complex from maternal ECG by a single thoracic channel. This method is based on the fixed-point convolution kernel compensation (FP-CKC) and recursive least square (RLS) method which is combined with particle swarm optimization (PSO). The rest of the paper is organized as follows: in the next section, information about the signals and formulation of methods used in this study is presented. Section 3 provides the results of the proposed method. The discussion is provided in Section 4, along with conclusions.

Section snippets

Materials and method

An overall block diagram of the proposed algorithm is shown in Fig. 1. As can be seen from the Figure, the signal should first pass from a sliding window. The zero centering and whitening should be applied on every window separately. Then, every sample of the window is used for the initial point of FP-CKC. The average of separated components is calculated and considered as the noise model or decomposed signal. The adaptive filter using the RLS algorithm and calculated noise model is applied on

Results on denoising

For measuring denoising quality, we used three types of validation parameters. The signal to error ratio (SER) is described in Eq. (13) [22] in which, xt and x˘(t) are original and reconstructed signal respectively and L is the length of the signal.SER(x,x˘)=t=0L-1x2(t)t=0L-1xt-x˘(t)2

Note that SER is calculated by using original ECG and as an instance when we calculate SER for noisy signal, it means that x˘(t) in Eq. (13) is set to noisy signal and the similarity of signals are evaluated.

State-of-the-art

One of the crucial issues is the importance of using an adaptive filter after FP-CKC. The modified versions of FP-CKC is used by other researches for blind source separation purposes, especially in neural decoding. However, in these applications, extracting firing times has more critical than finding exact shape of spikes and due to the number of recorded channels, the shape of spikes can be extracted by some simple post-processing. Furthermore, finding appropriate initial points for this

Declaration of Competing Interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

References (33)

  • W. Zong et al.

    A practical algorithm to reduce false critical ECG alarms using arterial blood pressure and/or photoplethysmogram waveforms

    Physiol. Meas.

    (2016)
  • S.-K. Teo

    Reducing false arrhythmia alarms in the ICU

    2015 Computing in Cardiology Conference (CinC)

    (2015)
  • J.A. Van Alsté et al.

    Removal of base-line wander and power-line interference from the ECG by an efficient FIR filter with a reduced number of taps

    IEEE Trans. Biomed. Eng.

    (1985)
  • S.L. Joshi et al.

    A survey on ECG signal denoising techniques

    Proceedings - 2013 International Conference on Communication Systems and Network Technologies, CSNT 2013

    (2013)
  • K. Bartyzel

    Adaptive Kuwahara filter

    Signal Image Video Process.

    (2016)
  • J. Dhiman et al.

    Comparison Between Adaptive Filter Algorithms (LMS, NLMS and RLS)

    (2013)
  • Cited by (27)

    • Fetal ECG extraction from maternal ECG using deeply supervised LinkNet++ model

      2023, Engineering Applications of Artificial Intelligence
    • Semi-supervised active transfer learning for fetal ECG arrhythmia detection

      2023, Computer Methods and Programs in Biomedicine Update
    View all citing articles on Scopus
    View full text