Identification and Classification of Cardiac Arrhythmia using Neural Network

Author Name:
Author Email:



This paper proposes a Neural Network classifier model for the automatic identification of cardiac arrhythmias. Our main aim is to have a better classification rate in a minimum time with simple algorithms in their implementation. The wavelet transform is used to extract the parameters which characterize different arrhythmic pathologies and thus constitute the bases of learning and generalization of the networks of neurons. A pre-processing by filtering adapted by the Butterworth Filter is applied to the signals before the extraction. Classification is done using Neural Network classifier further dimension reduction to reduce the overhead of the training is performed by PCA. The ECG signals used in this study are taken from the internationally known MIT-BIH Arrhythmia Database. The results obtained are satisfactory with average recognition rates of 95.7% in NN and 97.1 % for PCA-NN.


Irregular beats form two groups: the first comprises fibrillation and ventricular tachycardia which are dangerous and the ECG signal represents the electrical activity of the need for immediate therapy with a heart, is probably the diagnostic technique of the defibrillator; the second group concerns the most common cardiac pathologies. ECG is non-hazardous arrhythmias but which require a non-stationary randomized signal, structured preventive therapy of other problem. These disorders (Figure 1) by the succession of wave and cardiac forms are also classified according to the frequency and the interval of the (P, Q, R, S, and T). Any rhythm modification described by the number of cycles or morphological or temporal contractions of its events (or ventricular (depolarization of the ventricles) by chronic) constitutes a cardiac pathology. The minutes, marked on the ECG by the number of changes concerning the complex rhythm or frequency QRS are cardiac arrhythmias (fatal diseases). A number of works [1], [2], [3], [4], [5], [6], and [7] have been successfully performed for clinical and technological problem highlighting arrhythmias, but they defy by (means and approaches). They are


varied and the choice of types and number of parameters to approach them are also varied, significant that characterize the arrhythmia in question. It is possible to define the premature ventricles) in cardiac arrhythmias by the irregularity of the beats [6] [8] with cardiac arrhythmias or as a group of different parameters, approaches and results.

On the other hand, several classification methods have been proposed in the literature to classify the arrhythmic beats in normal or abnormal (arrhythmic) beats for some and others according to different rhythms (TV and BV) [4] , [9] and [10] compares three different procedures to characterize the rhythms to be classified by an LVQ network. But in most cases the essential parameters are measured around the QRS complex (width, shape and amplitude) and more particularly the detection of the R wave which is a dominant parameter [4], [6], [9]. Especially for atrial arrhythmias (atrial fibrillation) the P wave and the PR interval are privileged.



Figure 1: Characteristics of a heartbeat

This varied study allowed us to establish a number of significant parameters to classify the arrhythmias present in the QRS complex.

Inspired by several works cited, we propose an approach of automatic recognition to detect several arrhythmias. It involves methods of non-linear analysis by neural networks and tools of non-stationary analysis by the wavelet transform (WT) which have very interesting adaptation properties.

Currently the wavelet method has become a technique, par excellence, very effective to extract the parameters that characterize the arrhythmia in its various forms. Many works using wavelets have been successfully carried out, in particular the articles [11], [12] and [13] have amply demonstrated the contribution of the ECG signal. It makes it possible to produce the start and end times of waves and intervals and to filter the signal.

Nonlinear array analysis of artificial neurons (ANN) seems promising to provide indicators relevant to the identification of pathologies Heart disease. The literature is rich in works that have exploited the advantages of ANN to automatically detect or classify cardiac arrhythmias such as [4], [6] [14], [15], [16], [17], [18] and [19] where different classification approaches were discussed. But a classifier can be effective for an ECG and fail with another ECG or other pathology. This justifies the large number of works in this field. There will always be investigations because the slightest hope of improving performance saves a life. Previously Nath et al. [20] shown that for a particular study it is often difficult to identify the best classifier for lack of proper guide. Many comparative studies of classifiers are conducted. The combination of different approaches can be a powerful tool to improve the quality of classification algorithms. In particular, methods that take into account the a priori knowledge of the data have been more successful. This is why we propose vector quantization (VQ) learning algorithms whose essential tool is the codebook. By constructing the codebook, the VQ takes into account the specificities of the given application and thus makes it possible to exploit the information a priori. This is the combination of the SOM-LVQ algorithms and that of the RBF network. Each network (ANN) having learned the key characteristics of each arrhythmia, receives as input, in the test phase, a vector of parameters (defining a heartbeat) coming from the WT, discriminates in two or more classes of arrhythmias.

Our results are evaluated on the ECG recordings of MIT-BIH Data Arrhythmias, well known by ECG users, for detecting the following pathologies: Extra ventricular (V), Extra supraventricular systolic (S), Fibrillation Ventricular flutter (V), ventricular tachycardia (VT) and ventricular bradycardia (Br), which we will classify according to the rhythm and the frequency, the symbols are those used in MIT-BIH annotation.


The purpose of our work is to automatically classify detected cardiac arrhythmias on real ECG signals from the MIT-BIH database. The choice of nonlinear classification by artificial neural networks gives good performances because of their separation power from complex situations such as supraventricular arrhythmias. The performance of ANN also strongly depends on the quality of QRS extraction. The work that improved the extraction of the characteristics has had high results. The detection of the wavelet start and end of wavelets has been very effective for the QRS complex. This has guided our choice of approach. It is found that the efficiency of detection is nearly around 95.7% as the data is compared with MIT-BIH database.

The slightest improvement provided by the algorithms is a good result because it involves saving a life. We can say that our procedure also has important potentialities, but in order to be able to be the subject of a method of aid to the diagnosis (applicable clinically) one must first treat more ECGs, in particular for the rhythm, to do Vary the database and treat other cardiac pathologies. Theoretically, neural networks are optimum when the database is very large. It is necessary to find explanations for the contradictory results between the two networks.







  • Babloyantz, A. and lvanov, V.V., 1998. Neural networks in cardiac arrhythmias. In: Industrial Applications of Neural Networks, Pg: 403-417.


  • Silipo, R. and Marchesi, C., 1998. Artificial neural networks for automatic ECG analysis. IEEE transactions on signal processing, 46(5), Pg:1417-1425.


  • Gao, G.Q., 2003. Computerised detection and classification of five cardiac conditions (Doctoral dissertation, Auckland University of Technology).


  • Exarchos, T.P., Tsipouras, M.G., Exarchos, C.P., Papaloukas, C., Fotiadis, D.I. and Michalis, L.K., 2007. A methodology for the automated creation of fuzzy expert systems for ischaemic and arrhythmic beat classification based on a set of rules obtained by a decision tree. Artificial Intelligence in medicine, 40(3), Pg:187-200.


  • Tsipouras, M.G., Fotiadis, D.I. and Sideris, D., 2005. An arrhythmia classification system based on the RR-interval signal. Artificial intelligence in medicine, 33(3), Pg:237-250.


  • Shyu, L.Y., Wu, Y.H. and Hu, W., 2004. Using wavelet transform and fuzzy neural network for VPC detection from the Holter ECG. IEEE Transactions on Biomedical Engineering, 51(7), Pg:1269-1273.


  • Monzon, J.E. and Pisarello, M.I., 2005. Cardiac beat classification using a fuzzy inference system. In Engineering in Medicine and Biology Society, 2005. IEEE-EMBS 2005. 27th Annual International Conference of the IEEE, Pg: 5582-5584.


  • Saadi, S., Bettayeb, M., Guessoum, A. and Abdelhafidi, M., 2012. Artificial bees colony optimized neural network model for ECG signals classification. In Neural Information Processing Pg: 339-346. Springer Berlin/Heidelberg.


  • Assaleh, K. and Al-Nashash, H., 2005. A novel technique for the extraction of fetal ECG using polynomial networks. IEEE Transactions on Biomedical Engineering, 52(6), Pg:1148-1152.


  • Kristiansen, D.E., Husøy, J.H., Eftestøl, T., Erik, D. and John, K., 1995. Rhythm detection in ECG signals.


  • Senhadji, L., Carrault, G., Bellanger, J.J. and Passariello, G., 1995. Comparing wavelet transforms for recognizing cardiac patterns. IEEE Engineering in Medicine and Biology Magazine, 14(2), Pg:167-173.
  • Polat, K., Şahan, S. and Güneş, S., 2006. A new method to medical diagnosis: Artificial immune recognition system (AIRS) with fuzzy weighted pre-processing and application to ECG arrhythmia. Expert Systems with Applications, 31(2), Pg:264-269.


  • Jaylaxmi C. Mannurmath, Raveendra M., “MATLAB Based ECG Signal Classification”, International Journal of Science, Engineering and Technology Research (IJSETR), Vol. 3, Issue 7, July 2014.


  • Tang, L. Shu, “Classification of Electrocardiogram Signals with RS and Quantum Neural Networks”, International Journal of Multimedia and Ubiquitous Engineering, Vol. 9, No. 2, Pg: 363-372, March 2014.


  • Taiseer Mohammed Siddig, Mohmmed Ahmed Mohmmed, “A Study of ECG Signal Classification using Fuzzy Logic Control”, International Journal of Science and Research (IJSR), ISSN: 2319-7064, Vol. 3 Issue 2, February 2014.


  • Kavitha, T. Christopher, “A Study on ECG Signal Classification Techniques”, International Journal of Computer Applications, ISSN: 0975-8887, Vol. 86, No. 14, January 2014.


  • Sambhu, A. C. Umesh, “Automatic Classification of ECG Signals with Features Extracted Using Wavelet Transform and Support Vector Machines”, International Journal of Advanced Research in Electrical, Electronics and Instrumentation Engineering, ISSN: 2278-8875, Vol. 2, Issue 1, December 2013.


  • Devashree Joshi, Rajesh Ghongade, “Performance Analysis of Feature Extraction Schemes for ECG Signal Classification”, International Journal of Electrical, Electronics and Data Communication, ISSN: 2320-2084, Vol. 1, July 2013.


  • K. Srivastava, Dr. Devendra Prasad, “DWT – Based Feature Extraction from ECG Signal”, American Journal of Engineering Research (AJER), ISSN: 2320-0847, Vol. 02, Issue 03, Pg: 44-50, 2013.


  • Nath, S.S., Mishra, G., Kar, J., Chakraborty, S. and Dey, N., 2014, July. A survey of image classification methods and techniques. In Control, Instrumentation, Communication and Computational Technologies (ICCICCT), 2014 International Conference on Pg:554-557). IEEE.


Christos Stergiou and Dimitrios Siganos, “Neural Networks”, Report available at:



11 total views, 0 views today

About the author: tej