THE USE OF NEURAL NETWORKS FOR THE AUTOMATIC ANALYSIS OF ELECTROCARDIOGRAMS IN DIAGNOSIS OF CARDIOVASCULAR DISEASES

Objectives The aim of the work is to use neural networks to detect characteristic ECG signals that determine heart rhythm abnormalities and detect the corresponding heart disease. One of the most important factors for the timely provision of medical care is the rapid and accurate obtaining of information about the patient's state of health. The timeliness of diagnosis is often the main factor determining the patient's prognosis. While the accuracy of cardiological diagnosis has significantly increased in recent years due to the wide application of both instrumental and laboratory research methods, the percentage of diagnostic errors in cardiology remains high. Electrocardiography (ECG) consists ina non-invasive process of interpreting the electrical activity of the heart, allowing the speed and regularity of the heartbeat to be assessed. These data are then used to determine any heart defects or pathologies. However, automatic ECG analysis remains a challenging theoretical and practical task.

MethodsA MATLAB 8.6 (R2015b) Neural Network Toolbox was used to simulate artificial neural networksduring the design. A backpropagation algorithm was used for trainingthe neural network. The efficiency of the developed neural network model for ECG analysis was investigated using the MIT-BIH arrhythmia database.

Results The accuracy of detection and extraction of the components of the ECG signal shows that the developed neural network model can be successfully used to detect heart diseases among patients. The sensitivity of the model was 71%, with a specificity of 89%. The elaboration of existing and development of new methods for processing electrocardiographic data allow the problem of timely diagnosis and prevention of cardiovascular diseases to be solvedat early stages of their detection.

ConclusionThe accuracy of detection and extraction of the ECG signal components shows that the developed neural network model can be used to detect heart diseases among patients.

1. Vsemirnaya organizatsiya zdravookhraneniya [Elektronnyi resurs]: Serdechno-sosudistye zabolevaniya. URL: http://www.who.int/mediacentre/factsheets/fs317/ru (data obrashcheniya: 25.10.2017). [World Health Organization [Electronic resource]: Cardiovascular diseases. URL: http://www.who.int/mediacentre/factsheets/fs317/ru (access date: 25.10.2017). (in Russ.)]

2. Lee J.W., Lee G. Design of an adaptive filter with a dynamic structure for ECG signal. Processing. International Journal of Control, Automation, and Systems. 2005;3(1):137-142.

3. Ince T., Kiranyaz S., Gabbouj M. A generic and robust system for automated patient-specific classification of ECG signals. IEEE Transactions on Biomedical Engineering. 2009;56:1415-1426.

4. Lin C.H., Du Y.C., Chen T. Adaptive wavelet network for multiple cardiac arrhythmias recognition. Expert Systems with Applications. 2008;34:2601-2611.

5. Mustafaev A.G. Neirosetevaya model' prognozirovaniya urovnya glyukozy v krovi u bol'nykh sakharnym diabetom. Kibernetika i programmirovanie. 2016;3:1-5. [Mustafaev A.G. Neural network model for predicting the level of glucose in the blood in patients with diabetes mellitus. Cybernetics and programming. 2016;3:1-5. (in Russ.)]

6. Mustafaev A.G. Primenenie iskusstvennykh neironnykh setei dlya rannei diagnostiki zabolevaniya sakharnym diabetom. Kibernetika i programmirovanie. 2016;2:1-7. [Mustafaev A.G. The use of artificial neural networks for early diagnosis of diabetes mellitus. Cybernetics and programming. 2016;2:1-7. (in Russ.)]

7. Nikitin V.M., Lomakin V.V., Anokhin D.A., Kaidalova I.K., Ivanov I.I. Informatsionnaya sistema podderzhki prinyatiya reshenii slabostrukturirovannykh zadach v kardiologii. Nauchnye vedomosti BelGU. Seriya “Istoriya. Politologiya. Ekonomika. Informatika”. 2010;19(90):112–119. [Nikitin V.M., Lomakin V.V., Anokhin D.A., Kaidalova I.K., Ivanov I.I. Information system for decision support of weakly structured tasks in cardiology. Belgorod State University Scientific bulletin. Series “History. Political science. Economics. Information technologies”. 2010;19(90):112–119. (in Russ.)]

8. Fainzil'berg L.S., Lebedushko T.Yu. Issledovanie diagnosticheskoi tsennosti fazovykh portretov EKG po dannym spetsializirovannykh baz. Kibernetika i vychislitel'naya tekhnika. 2012;169:34-50. [Fainzil'berg L.S., Lebedushko T.Yu. Study of the diagnostic value of ECG phase portraits according to the data of specialised bases. Cybernetics and Computer Engineering. 2012;169:34-50. (in Russ.)]

9. Kalinichenko A.N. Otsenka razdelyayushchei sposobnosti metodov klassifikatsii form EKG. Izvestiya SPbGETU “LETI”. Ser. Biotekhnicheskie sistemy v meditsine i ekologii. 2006;1:21-30. [Kalinichenko A.N. Evaluation of the separating ability of ECG form classification methods. Izvestiya SPbGETU “LETI”. Ser. Biotekhnicheskie sistemy v meditsine i ekologii. 2006;1:21-30. (in Russ.)]

10. Portal iskusstvennogo intellekta [Elektronnyi resurs]: Iskusstvennyi neiron. Struktura iskusstvennogo neirona. URL: http://neuronus.com/theory/151-iskusstvennyj-nejron.html. (data obrashcheniya: 25.12.2017). [Portal of artificial intelligence [Electronic resource]: Artificial neuron. Structure of an artificial neuron. URL: http://neuronus.com/theory/151-iskusstvennyj-nejron.html. (access date: 25.12.2017). (in Russ.)]

11. Guangying Y., Yue C. The study of electrocardiograph based on radial basis function neural network. Intelligent Information Technology and Security Informatics (IITSI). 2010:143-145.

12. Gao D., Madden M. Bayesian ANN classifier for ECG arrhythmias diagnostic system. Proceeding of IEEE International Joint Conference on Neural Network. 2003. P. 2454-2459.

13. Baza dannykh aritmii Massachusetskogo tekhnologicheskogo instituta [Elektronnyi resurs]: Opisanie. URL: http://ecg.mit.edu/ (data obrashcheniya: 20.09.2017). [The database of arrhythmias of the Massachusetts Institute of Technology [Electronic resource]: Description. URL: http://ecg.mit.edu/ (access date: 20.09.2017). (in Russ.)]

14. Gabriel' Khan M. Bystryi analiz EKG. M.: Binom; 2009. 408 s. [Gabriel' Khan M. Quick analysis of the ECG. M.: Binom; 2009. 408 p. (in Russ.)]

15. Trenn S. Multilayer perceptrons: approximation order and necessary number of hidden units. IEEE Transactions on Neural Networks. 2008;19(5):836–844, 2008.

16. Xu S., Chen L. A novel approach for determining the optimal number of hidden layer neurons for FNN's and its application in data mining. Proceedings of the 5th International Conference on Information Technology and Applications (ICITA '08). 2008. P. 683–686.

17. Sun J. Learning algorithm and hidden node selection scheme for local coupled feedforward neural network classifier. Neurocomputing. 2012;79:158–163.

18. Nazarov A.V., Loskutov A.I. Neirosetevye algoritmy prognozirovaniya i optimizatsii sistem. SPb: Nauka i tekhnika; 2003. 380 s. [Nazarov A.V., Loskutov A.I. Neural network algorithms for forecasting and optimisation of systems. SPb: Nauka i tekhnika; 2003. 380 p. (in Russ.)]

19. Masters T. Practical Neural Network Recipes in C++. Academic Press; 1993. 504 p.

20. Salekh M.A., Isakov R.V. Otsenka effektivnosti primeneniya iskusstvennykh neironnykh setei dlya analiza segmentirovannykh elektrokardiokompleksov. Biomeditsinskaya radioelektronika. 2012; 6:21-27. [Salekh M.A., Isakov R.V. Eeffectiveness evaluation of the application of artificial neural networks for the analysis of segmented electrocardiocomplexes. Biomedical radioelectronics. 2012;6:21-27. (in Russ.)]

21. Galushkin A.I. Teoriya neironnykh setei. M.: IPRZhR; 2000. 348 s. [Galushkin A.I. The theory of neural networks. M.: IPRZhR; 2000. 348 p. (in Russ.)]