ReviewWavelet-based electrocardiogram signal compression methods and their performances: A prospective review
Introduction
Recent developments in digital technologies, sensors, efficient signal processing tools and wireless communication technologies enables to store and transmit biomedical signals for diagnosing patients diseases. Nowadays cardiovascular disease (CVD) is one of the most widespread health problems with unpredictable and life-threatening consequences in developing and developed countries in the world. The electrocardiogram (ECG) signals are widely used for detecting different kinds of heart diseases [1], [2], [3], [4], [5], [6], [7], [8]. For effectively detecting and diagnosing CVD problems, long-term multi-channel ECG signals are recorded with different sampling rates and bit resolutions. The ECG signals, for example, received from recording systems such as 12-lead ECG, the vector cardiography (VCG) high-resolution ECG, exercise ECG are digitized at the sampling rate ranging from 100 to 1000 Hz with resolution between 8 and 12 bit per sample [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18]. The amplitude of the recorded ECG signal on the skin is from 0.1 mV to 5 mV [1], [2], [3], [4], [5], [6], [7], [8], [9], [10]. The recording of the electrical field generated by the His and Purkinje activities produces a signal in the ECG with an amplitude range of about 1–10 μV which is useful for identification of conduction abnormalities [2]. The frequency extends from 0.05 Hz to 130 Hz [3], [10], [11], [12]. In ECG signal, notches and slurs may be superimposed on the slowly varying QRS complexes. The recommendations of the committee on electrocardiography of the American heart association suggest a conversion rate of 500 Hz with a 9-bit resolution [1]. The amount of ECG data depends upon the sampling frequency, sample width, the number of leads and record duration. Thus, the long-term ECG monitoring system generates large amount of cardiac data. Recently, miniaturized wearable ECG recorder enables both recording and transmission of ECG data via well-established telecommunication networks to the specialist diagnostic center [13], [14], [15]. In practice, sample rates from 100 Hz to 1000 Hz are used with 8–16 bit resolution per sample [1], [10], [16], [17], [18]. The data rate has become approximately 11–22 Mbits/h/lead. For the CVD analysis, multi-channel ECG signals are continuously recorded for 12–72 h to monitor ischemia, ventricular and supra-ventricular dysrhythmias, conduction abnormalities, QT interval and heart rate variability [2], [3], [4], [5], [6], [7], [8]. Hence, ECG record management system and telecardiology application requires reliable and efficient compression techniques for storage and continuous transmission purposes [34].
In general, single-channel and multi-channel ECG signal recording systems are used for effective analysis and diagnosis of cardiovascular diseases. Under long-term multi-channel recording conditions, ECG signals may exhibit three types of signal correlations: the intra-beat, the inter-beat and the inter-channel/lead. Fig. 1, Fig. 2, Fig. 3 illustrate different types of signal correlations. The intra-beat correlation represents the correlation between the successive samples in an ECG cycle. The inter-beat correlation represents the correlation between successive beats in a single-channel ECG signal. The correlation that exists between the signals from different channels is termed as inter-channel correlation. For effective storage and transmission purposes, there is a need for ECG data compression without introducing clinical distortion. The data compression is a method which is mainly developed: to reduce the channel bandwidth required for sending a given amount of information in a given time; to reduce the time required for transmitting a given amount of information in a given channel capacity [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37]. In general, the long-term multi-channel ECG signals have three correlations including inter-sample correlation within ECG beat, inter-beat correlation and inter-channel correlation. The ECG signals may contain long or short isoelectric regions between two local events or waves. The signal redundancy can be exploited when successive ECG samples are statistically dependent and the quantized ECG sample amplitudes occur with unequal probability [19], [29], [34]. In most ECG compression methods, the signal irrelevancy and redundancy properties in the time-domain, frequency-domain and time-frequency domain are exploited by using different signal processing techniques.
Section snippets
Compression efficiency
In most methods, the compression efficiency is evaluated in terms of the following performance measures: the sample reduction ratio (SRR), the compression ratio (CR), the compressed data rate (CDR) and the decoding rate (DR) [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35]. The sample reduction ratio (SRR) is defined as
The compression measure
Classification of ECG compression methods
A digital information source is said to possess redundancy if either [20]: the symbols are not equally likely, or the symbols are not statistically independent. All data compression methods seek to minimize data storage by reducing the redundancy [21]. Redundancy in a set of discrete samples exists when signal samples are statistically dependent [27]. Data compression methods, which are successfully used for speech, image and video signals, can be employed for ECG signals. ECG data compression
Wavelet transform
In many signal processing applications, mathematical transformations are applied to a signal in order to extract information which may not be readily available from the original signal. A wavelet based signal processing technique is an effective tool for nonstationary ECG signal analysis and characterization of local waves (P, T and QRS complex morphologies). The details of design of wavelets and wavelet transforms can be found in [176], [177], [178], [179], [180], [181], [182], [183]. In this
One-dimensional wavelet based ECG compression methods
In ECG data compression, the wavelet and the wavelet packet transform are used to exploit the redundancy in the signal. After the signal is transformed into the wavelet domain, many coefficients are so small that no significant information is lost in the signal by setting these coefficients to zero [142]. Many 1-D and 2-D wavelet based compression methods are reported in literature. The 2-D wavelet-based ECG compression method achieved high compression ratio with low reconstruction error at the
Performance evaluation of ECG compression methods
In this section, the compression issues in the reported methods are described and investigated with different sets of experiments. In [136], [138], [172], various tests were carried out using the mita database and the PRD measurement criteria to evaluate the quality of the compressed signal and to select optimal coding parameters. In the next subsections, experimental results are presented for different signal processing techniques used in the ECG compression methods.
Discussion and concluding remarks
From the literature survey, many ECG compression methods have been proposed by exploiting the short-term correlation between adjacent samples, the considerable similarity between adjacent ECG beats (long-term correlation) and between the intra-lead ECG beats. In recent years, the wavelet transform has proven to be an effective tool in noise reduction, clinical features extraction and particularly ECG data compression because of its interesting properties such as time-frequency localization,
References (204)
- et al.
Compact digital storage of ECGs
Comput. Programs Biomed.
(1979) - et al.
Third order polynomial-Its use in data compression
Signal Process.
(1988) - et al.
Fast ECG data compression algorithms suitable for microprocessor systems
J. Biomed. Eng.
(1992) AZTDIS-a two-phase real-time ECG data compressor
J. Biomed. Eng.
(1993)- et al.
ECG data compression via cubic-splines and scan-along polygonal approximation
Signal Process.
(1994) - et al.
ECG data compression by spline approximation
Signal Process.
(1997) Karhunen-loeve representation of ECG data
J. Electrocardiol.
(1992)- et al.
ECG signal compression plus noise filtering with truncated orthogonal expansions
Signal Process.
(1999) - et al.
A fast Walsh transfrom-based data compression multi-microprocessor system: application to ECG signals
Math. Comput. Simul.
(1985) ECG data compression using adaptive Fourier coefficients estimation
Med. Eng. Phys.
(1994)
A dynamic fourier series for the compression of ECG using FFT and adaptive coefficient estimation
Med. Eng. Phys.
ECG component delineation by Prony's method
Signal Process.
ECG data compression by modeling
Comput. Biomed. Res.
Compression of ECG signals by optimized quantization of discrete cosine transform coefficients
Med. Eng. Phys.
Constrained ECG compression algorithm using the block-based discrete cosine transform
Digit. Signal Process.
Recommendations for standardization of leads and of specifications for instruments in electrocardiography and vectorcardiography
Circulation
ABC of Clinical Electrocardiography
Biomedical Digital Signal Processing
Biomedical Signal Analysis: A Case-study Approach
Research in electrocardiography and magnetocardiography
Proc. IEEE
Origin of the electrocardiogram
IEEE Eng. Med. Biol. Mag.
ECGs for the Emergency Physician
Blueprints in Cardiology
Cardiovascular Signal Compression Using a New Wavelet Energy based Diagnostic Distortion Measure
The use of the SPSA method in ECG analysis
IEEE Trans. Biomed. Eng.
Estimation of QRS complex power spectra for design of a QRS filter
IEEE Trans. Biomed. Eng.
High-frequency electrocardiogram analyzer
IEEE Trans. Biomed. Eng.
Wavelet-based ECG and PCG signals compression technique for mobile telemedicine
Wavelet-threshold based ECG compression with smooth retrieved quality for telecardiology
Automated cardiac event change detection for continuous remote patient monitoring Devices
The MIT-BIH arrhythmia database in CD-ROM and software for use with it
The impact of the MIT-BIH arrhythmia database
IEEE Eng. Med. Biol. Mag.
PhysioNet: A web-based resource for study of physiologic signals
IEEE Eng. Med. Biol. Mag.
ECG data compression for tapeless ambulatory monitors
Comput. Cardiol.
Data compression
ACM Comput. Surv.
Digital Coding of Waveforms
Introduction to Data Compression
Principles of quantization
IEEE Trans. Circuits Syst.
Speech coding based upon vector quantization
IEEE Trans. Acoust. Speech Signal Process. ASSP-28
Vector quantization
IEEE Acoust. Speech Signal Process. Mag.
Image coding using vector quantization: a review
IEEE Trans. Commun.
Redundancy reduction-a practical method of data compression
Proc. IEEE
A new data reduction algorithm for real time ECG analysis
IEEE Trans. Biomed. Eng.
ECG data compression techniques-a unified approach
IEEE Trans. Biomed. Eng.
Adaptive data compression
Proc. IEEE
Predictive coding
IEEE Trans. Inform. Theory
Compact digital coding of electrocardiographic data
Compression of the ECG by prediction or interpolation and entropy encoding
IEEE Trans. Biomed. Eng.
Redundancy reduction, the key to adaptive telemetry
Data redundancy reduction for biomedical telemetry
Cited by (114)
Lightweight electrocardiogram signal compression
2023, Biomedical Signal Processing and ControlFrom signal to image: An effective preprocessing to enable deep learning-based classification of ECG
2023, Materials Today: ProceedingsA novel ECG signal compression using wavelet and discrete anamorphic stretch transforms
2022, Biomedical Signal Processing and Control