FIELD: medicine.
SUBSTANCE: invention refers to medicine, particularly to cardiology. Clinico-anamnestic data and results of patient echocardiography are used to construct a mathematical model of artificial neural networks. In this case, the neural network is represented by a multilayer perceptron consisting of 25 neurons of the input layer, representing the following factors: sex, age, body mass index, smoking, associated arterial hypertension and diabetes mellitus, presence of aneurysm and thrombus in the left ventricular cavity, size of the left and right ventricles, the left atrium, the interventricular septum, aortic root diameter, presence of mitral regurgitation, size of the asynergia and left ventricular ejection fraction, functional class of chronic heart failure. In addition, neural network consisting of a hidden layer of six neurons with an activation function is a hyperbolic tangent in the form of 
and an output layer consisting of two neurons: ||HGVA(1)|| – presence of arrhythmia and ||HGVA(0)|| – no arrhythmia, with normalization of the values by Softmax function. If the value ||HGVA(1)|| is greater than or equal to ||HGVA(0)||, a conclusion is drawn about the risk of development of ventricular arrhythmias of high grades, and if the value ||HGVA(1)|| is less than ||HGVA(0)||, the development of arrhythmias is not predicted.
EFFECT: method makes it possible to increase the accuracy of predicting gastric arrhythmias of high grades directed to coronary angiography, and to shorten the time of examination.
1 cl, 2 ex, 3 tbl
