R.V. Isakov, M.A. Saleh, L.T. Sushkova
Currently, there is positive dynamics of growth in the number of cardiovascular diseases. Therefore, strongly increases importance of hardware and software information and analytical systems to conduct automated rapid population study to identify deviations from the norm.For the detection of diseases of the cardiovascular system developed models of artificial neural networks. Application of neural network analysis in clinical practice improves the accuracy of diagnosis of diseases of the cardiovascular system.
This paper presents the results of research to create a hardware-software complex neural network ECS analysis.
For ECS pretreatment was chosen an approach based on segmentation electrocardiocomplexes on 3 key areas that are responsible for depolarization of the atria, ventricles depolarization and repolarization of the ventricles.
Abnormalities in the signal in each segment are determined by using an artificial neural network, which is less sensitive to noise arising in the signal and is capable of generalization. Research has shown good results of neural networks like multilayer perceptron for analysis of ECG.
Developed structural diagram the hardware-software complex neural network analysis ECS consists of 12 units: amplification and analog-digital conversion, program unit of signal recording, processing and visualization in real time, unit of signal pretreatment and selection of images, analysis units of atrial, ventricular, and repolarization segments of ECG, analysis unit of scattergram and histogram, decision-making unit, report generator.
For the hardware part of the complex can be applied to practically any digital cardiograph, providing registration of standard leads, the sampling frequency not less than 500 Hz and resolution of at least 12 bits.
This complex can be used in computer-aided analysis of the functional state of the cardiovascular system during mass rapid – studies with a view to providing "risk groups", as well as in programs of the daily ECG recording automated interpretation.
Currently under development and improvement of the hardware-software complex, and develop new neural network modules that increase the functionality of the system.