R.V. Isakov, M.A. Al-Mabruk, J.A. Lukyanova, L.T. Sushkova
Presently cardiovascular diseases are main reason of man population grown death rate in different countries, thus its decline is possible only by an effective prophylaxis. The purpose of this research is questions consideration of neuron networks application expedience and features in the automated determination of pathological changes to electric activity of heart.
Among the different structures of neuron networks, most known is a multi-layered structure (multi-layered perceptron). Development of neural networks methods enabled the use them as an instrument of scientific researches, by which it is possible to study objects and phenomena.
Comparison of multi-layered perceptron and modular construction of neuron network organization was in-process produced, as variants of the automated analysis of elecrtocardiosignal system construction.
Two databases were formed: base of heart pathologies and arrhythmias. They were grouped in a few basic classes and part on two independent databases: teaching and test.
Researches carried out by the frequent teaching of artificial neuron networks of the hidden layer different volume. After teaching every network passed testing on the independent database.
Researches shown: good sensitiveness to pathologies, a low error level and possibility of unlimited extension of analyzed pathologies number. It allows to consider that modular construction, probably, is an optimum choice for the decision of electrocardiosignal analysis task.