FIELD: medicine.
SUBSTANCE: (a) based on the clinical anamnestic data, results of laboratory studies, echocardiography and Holter monitoring data of the patient, a mathematical model of artificial neural networks is built in the form of multi-layer perceptron, consisting of 6 neurons in the input layer (age, heart rate, systolic blood pressure on admission to the hospital, the maximum value of CPK, final diastolic size of the left ventricle and the size of myocardial asynergia), a hidden layer of eight neurons with a hyperbolic tangent activation function in the following form: f(z)=(ez-e-z)/(ez+e-z), an output layer, consisting of two neurons: 
- presence of risk of paroxysmal ventricular tachycardia development and 
- absence risk of paroxysmal ventricular tachycardia development, with normalization of the output layer values by Softmax function. When the value of 
is greater than or equal to 
, a risk of paroxysmal ventricular tachycardia development is concluded, and when the value of 
is less than 
, ventricular tachycardia development is not predicted.
EFFECT: method allows to accurately predict the development of the paroxysmal ventricular tachycardia, which may contribute to more effective prevention of sudden cardiac death and other life-threatening complications in patients in the early day of myocardial infarction.
3 tbl, 2 ex
