Improvement of electrocardiogram (ECG) signal classification with support vector machine and particle swarm optimization (PSO-SVM)

Number of pages: 70 File Format: word File Code: 32171
Year: 2014 University Degree: Master's degree Category: Electronic Engineering
  • Part of the Content
  • Contents & Resources
  • Summary of Improvement of electrocardiogram (ECG) signal classification with support vector machine and particle swarm optimization (PSO-SVM)

    Dissertation for Master's Degree

    Department of Medical Engineering

    Abstract

    Cardiac arrhythmias are one of the heart diseases that should be paid attention to in patients admitted to the intensive care unit. Smartening the process of accurate diagnosis of heart diseases is a problem that has attracted the attention of researchers for years. In this research, an efficient method for selecting the appropriate features extracted from the ECG signal, based on the Binary Cuckoo Algorithm (BCOA), is presented. The extracted features include temporal features, AR and wavelet coefficients, the number of these features is reduced using the mRMR or PCA operator. BCOA forms sets of features and always seeks to find a suitable set of all features. The evaluation of this set of features selected by BCOA is investigated by applying to SVM classifier. Then the PSO algorithm is applied to optimize the parameters of SVM. With the help of computer simulation, the overall accuracy of the system for identifying 6 types of heart rhythms was 97.98%, which shows the optimal performance of the proposed method when comparing the accuracy with previous researches.

    Introduction

    A signal is a function of one or more independent variables that contains information about a physical or biological phenomenon. Living organisms, from cells to body organs, produce signals of biological origin. These signals are electrical, mechanical or chemical. Electrical signals are the result of depolarization of nerve cells or heart muscle. The sound produced by the heart valves is an example of a mechanical signal. These biological signals or vital signals are used for medical diagnosis and bio-medical research.

    The vital signals on the surface of the body reflect the internal state and electrical activity of the body. Therefore, it provides information about internal organs using non-invasive measurement. Electrocardiograms are used by cardiologists for diagnostic purposes and provide key information about the electrical activity of the ECG [1]. Therefore, with the constant display of this signal, the changes in the heart's electrical activity can be seen over time, and these changes include very key information for doctors [1].

    1-2- Problem definition

    The heart is one of the most important organs of the body that is responsible for pumping blood in the cardiovascular system. If the heart functions out of its natural order (rhythm), the blood circulation is not done well and this can cause serious risks for the person, therefore, correct and timely diagnosis of cardiac arrhythmias is very important. One of the well-known ways to detect these arrhythmias in time is to examine the electrical activities of the heart using electrocardiographic signals or ECG for short. Significant changes in the heart structure of patients and its beats can be detected using these signals [2]. In recent years, the automatic classification of electrocardiogram signals has attracted the attention of medical engineers. Through these signals, a cardiologist will have useful information about the heart's rhythm and function. Therefore, its analysis shows an effective way to identify and treat all kinds of heart diseases [3].

    To design an intelligent system for detecting cardiac arrhythmias from electrocardiographic signals, it is necessary to first extract appropriate features from these signals. Considering that the wavelet coefficients are able to describe the time-frequency information of the signal together, one of the choices will be to extract features from an electrocardiographic signal. In this regard, the number of levels of decomposition and the type of wavelet should be determined. Also, the results of previous researches have shown that for feature extraction from electrocardiographic signals, Dabichz and Haar family are much more suitable compared to other wavelets [4].The doctor's diagnosis is based on temporal and morphological information extracted from the electrocardiographic signal. While sometimes wavelet analysis on heart signals alone is not enough for classification, and for this reason it is necessary to use other features in heart signals to classify heart diseases. In order to more fully describe the electrocardiographic signals, in addition to the wavelet characteristics, temporal characteristics are also used. [4[.

    1-3- Necessity and importance of research

             Since ECG enables the doctor to record the electrical activity of the heart, it can be used to diagnose heart diseases. Automatic computer analysis is used to eliminate human error and to use existing databases in accurate and quick diagnosis of diseases. Therefore, in this research, an attempt has been made to automatically diagnose heart diseases, which will eliminate human errors in the diagnosis of diseases in the foreseeable future. The purpose of this research is to provide a suitable method for the automatic diagnosis of 5 important heart diseases, including RBBB[2], LBBB[3], PVC[4], APC[5] and P[6] failures. They will be done to select suitable signals and windowing on them. Then suitable features are extracted and classification is done based on these features. The above steps will be done using MATLAB software.

    1-5- Definition of concepts

    Electrical signal of the heart:

    The propagation of the action potential in the heart creates a current. This current in turn produces an electric field that can be obtained using a differential voltage measurement system in a very weak form on the surface of the body. The signal measured in this way, when captured by electrodes at standard locations, is known as an electrocardiogram, or ECG for short. The normal ECG signal is in the range of ±2 mv, and to record it, a device with a bandwidth of 0.5 to 15 Hz is needed. In other words, ECG is a graphic representation of the heart's activity in the form of an electrical signal recorded over a period of time[5]. Adequate blood supply to body tissues requires a sufficient number of heart beats, and the timing and sequence of heart muscle contractions must be carefully coordinated. The heart's natural pacemaker is the SA node, a microscopic group of specialized cardiac electrical cells located at the top of the right atrium. Following an electrical stimulation by the "sino-atrial node", a heartbeat is generated. This stimulation is transmitted to the muscle tissue cells of the heart walls through specific pathways. This stimulation first contracts the upper chambers of the heart, the atria, and pushes the blood into the ventricles. The stimulation is then transmitted to another area of ??electrical cells called the AV node, located above the ventricles. This node acts as a delay station in the stimulation pathway and allows the atria to fully empty. After a short period of time, the stimulation enters the ventricles through the branching paths and leads to their contraction. In this paper, an efficient method to select appropriate extracted features from the electrocardiogram (ECG) signal is presented, which is based on the Binary Cuckoo Optimization Algorithm (BCOA). The extracted features include Time features, Autoregressive (AR) and Wavelet; and the number of these features is reduced by using the Minimum Redundancy Maximum Relevance (mRMR) method or Principal Component Analysis (PCA) functions. Trying to find a suitable set of all features, BCOA creates sets of features

  • Contents & References of Improvement of electrocardiogram (ECG) signal classification with support vector machine and particle swarm optimization (PSO-SVM)

    List:

     

    Chapter One                      Introduction. 1

    1-1- Introduction. 2

    1-2- Definition of the problem. 2

    1-3- Necessity and importance of research. 3

    1-4- Research method. 3

    1-5- Definition of concepts. 4

    The electrical signal of the heart: 4

    The action potential of the heart muscle. 5

    Rest stage: 5

    Depolarization stage: 5

    Repolarization stage: 5

    P wave: 6

    QRS curve: 6

    T wave: 6

    ST piece: 6

    QT interval: 6

    Heart beat diseases: 6

    The second chapter Research background. 2

    2-1- Introduction. 10

    Introducing the database: 10

    2-2- ECG signal classification using wavelet and neural network. 10

    2-3- ECG signal classification using wavelet and morphological properties and neural network. 11

    2-4- ECG signal classification using wavelet transform and fuzzy neural network. 11

    2-5- ECG signal classification using violet transform and artificial neural network and bird algorithm. 12

    2-6- Classification of cardiac arrhythmias using SVM. 12

    2-7- Classification of atrioventricular arrhythmia. 12

    2-8- Electrocardiogram signal classification with support vector machine classifier and PSO algorithm. 13

    2-9- Classification of cardiac arrhythmias using PSO. 13

    2-10- Combined approach in cancer classification. 14

    2-11- Classification of cardiac arrhythmias based on wavelet transform and SVM. 14

    2-12- ECG signal classification using morphological properties. 14

    2-13- feature selection using binary cuckoo algorithm. 14

    2-14- feature selection using the cuckoo algorithm. 15

    Chapter Three Introduction of ECG signal processing algorithms and methods. 10

    3-1- Introduction. 17

    3-2- Wavelet analysis. 17

    3-2-1- Continuous wave transformation (CWT) 18

    3-2-2- Discrete wavelet transformation. 18

    3-3-2-2- multi-level decomposition. 18

    3-2-4- Selecting the mother wavelet. 19

    3-2-4- Features extracted from Violet. 21

    3-3- time characteristic. 21

    3-4- Feature extraction with autoregression model (AR) 22

    3-5- Feature selection strategy. 22

    3-6-Principal component analysis (PCA) 23

    3-7- Maximum dependence and minimum redundancy (mRMR) method 24

    3-8- COA cuckoo algorithm. 26

    3-8-2- Details of the cuckoo optimization algorithm. 27

    3-8-2-1- Producing the primary habitats of cuckoos (initial population of candidate answers) 29

    3-8-2-2- Method of cuckoos for laying eggs. 30

    3-8-2-3- Migration of cuckoos 30

    3-8-2-4- Destroying cuckoos located in inappropriate areas. 32

    3-8-2-5- algorithm convergence. 32

    3-9- Binary discretization of the cuckoo algorithm. 33

    3-10- Support vector machine (SVM) 33

    3-11- Particle optimization algorithm (PSO) 35

    3-11-1- Inertia weight. 36

    3-12- Overview of ECG signal classification system. 38

    Chapter IV Proposed ECG signal classification method. 17

    4-1- Introduction. 40

    4-2- ECG signal preprocessing. 41

    4-2-1- Signal shift to background deviation. 42

    4-2-2- Removing the average value of the signal. 42

    4-2-3- Removing noise caused by city electricity. 43

    4-2-4- Signal smoothing. 43

    4-2-5- signal windowing. 43

    4-2-6- Correlation test and removal of uncorrelated beats. 44

    4-2-7- Selection of training and test data. 44

    4-3- signal characteristics. 47

    4-3-1- feature extraction. 47

    4-3-1-1- Time feature. 47

    4-3-1-2- wavelet feature. 47

    4-3-1-3- AR feature. 47

    4-3-1-4- Identification of important signal points using PCA. 48

    4-3-2-combination and integration of features 48

    4-3-2-1- feature selection with PCA. 48

    4-3-2-2- feature selection with mRMR. 49

    4-3-2-3- feature selection using the cuckoo algorithm. 49

    4-4- Classification using SVM. 51

    Chapter Five Conclusion. 55

    5-1- Introduction. 56

    5-2- Comparison and conclusion. 56

    5-4- Submitting an offer. 57

    Resources: 58

     

     

    Source:

     

     

    ]1 [Mozen,57

    Sources: 58

     

     

    Source:

     

     

    ]1 [Mozen, Iman. Ahmadzadeh, Mohammad Reza. (2008), Designing a recognition system to classify heart signals based on wavelet transform and probabilistic neural networks, Majlesi scientific research journal, third year, third issue, pp. 31-37. [p>

    ]2[ Qobadi Mohebi, Sahar. (2008), Diagnosis and classification of cardiac arrhythmias using SVM method, 13th Iranian Electrical Engineering Student Conference.

    [3] Melgani, F. And Bazi, Y. (2008), Classification of Electrocardiogram Signals with Support Vector Machines and Particle Swarm Optimization, IEEE Transactions on Information Technology in Biomedicine, VOL. 12, No. 5, pp. 667-677.

    ]4 [Makhlasi, Omid. Mehrshad, Naser. Razavi, Seyyed Mohammad. (2017), using hybrid structures of neural networks to detect cardiac arrhythmias by integrating wavelet and time features, intelligent systems in electrical engineering, second year, first issue. [5] Khazaei, Ali. (2009), classification of heart diseases, Noshirvani University of Technology master's thesis.

    [6] R.G. Mark, and G. B. Moody, MIT-BIH Arrhythmia Database 1997 [Online]. Available: http://ecg.mit.edu/dbinfo.html.

    [7] Ghorbanian, P. Ghaffari, A. Jalali, A. Nataraj, C. (2010) Heart arrhythmia detection using continuous wavelet transform and principal component analysis with neural network classifier. Computing in Cardiology IEEE Conference, pp 669-672.

    [8] Guyton, Arthur; Medical Physiology, Sepehri and others, Tehran, Andishe Rafi, 2017, third edition, p. 56. [9] L.Y. Shyu, Y.H. Wu, W.C. Hu, (2004) Using wavelet transform and fuzzy neural network for VPC detection from the Holter ECG, IEEE Trans. Biomed. Eng. 51, 1269–1273.

    [10] T. Ince, S. Kiranyaz, and M. Gabbouj, (2009) A Genetic and Robust System for Automated Patient-Specific Classification of Electrocardiogram Signals, IEEE Trans. Biomed. Eng. 56, 1415-1426.

    [11] Testing and Reporting Performance Results of Cardiac Rhythm and ST Segment Measurement Algorithms, ANSI/AAMI EC57:1998 standard, Association for the Advancement of Medical Instrumentation, 1998.

    ]12[ Ketayoun Shojayan, Rasool Amirfatahi, (2017) Identification of acute atrial-ventricular arrhythmias. Basis of segmentation and recognition, Majlisi electrical engineering, second year, first issue.

    [13] I. Guyon, A. Elisseeff(2003), "An introduction to variable and feature selection", Journal of Machine Learning Research, Vol. 3, pp. 1157–1182.

    [14] Khazaee, A. (2013), Heart Beat Classification Using Particle Swarm Optimization, I.J. Intelligent Systems and Applications, Vol.5, No. 6, pp. 25-33.

    [15] Y. Peng, W. Li, Y. Liu, (2006) A hybrid approach for biomarker discovery from microarray gene expression data for cancer classification, Cancer Informatics, Vol. 2, pp.301–311.

    ]16 [Mimeri chelate; Affection; Shirazi; Jalil; (2011); classification of cardiac arrhythmias based on wavelet transform and SVM; 4th Iran Electrical and Electronics Engineering Conference. [17] de Chazal F. and R. B. Reilly, (2006) A patient adapting heart beat classifier using ECG morphology and heartbeat interval features, IEEE Trans. Biomed. Eng., vol. 53, no. 12, pp. 2535–2543.

    [18] L. A. M. Pereira, D. Rodrigues, T. N. S. Almeida, C. C. O. Ramos, A. N. Souza, X.-S. Yang, J. P. Papa, (2013) BCS: A Binary Cuckoo Search algorithm for feature selection, Studies in Computational Intelligence, Vol 516, pp 141-154.

    [19] D Rodrigues, Pereira, L.A.M. ; Almeida, T.N.S. ; Pope, J.P. ; Souza, A.N. ; Ramos, C.C.O. ; Xin-She Yang, (2013) BCS: A binary cuckoo search algorithm for feature selection, Circuits and Systems (ISCAS), 2013 IEEE International Symposium, 465-468.

    [20] Ghanbari, Karami-Mollaei, (2006) A new approach for speech enhancement based on the adaptive thresholding of the wavelet packets, speech communication, Elsevier.

    [21] Burrus, C. Sidney, Gopinath, Ramesh ?., and Guo, Haitao. (1998). Introduction to Wavelets and Wavelet Transforms: A Primer. Upper Saddle River, New Jersey: Prentice-Hall.

    [22] R. N. Khushaba, A. Al-Jumaily, and A

Improvement of electrocardiogram (ECG) signal classification with support vector machine and particle swarm optimization (PSO-SVM)