Session S42.2

Neural Network Model for the Prediction of the Evolution of the First Appearance of Stenocardia

OV Melnik*

Ryazan State Radioengineering University
Ryazan, Russia

The first time appeared stenocardia (FTAS) is the first episode of pain registered of patient with myocardial ischemia. The prognosis for a disease for patients with the FTAS greatly differs so necessary to reveal and take under intensive supervision patients with the high risk of the myocardial infarction and ventricular fibrillation.
The purpose of the work is development of the automatic system, allowing to construct the prognosis of current FTAS on base of the set of risk factor and to reveal patients with high risk of the myocardial infarction and sudden cardiac death.
In study are included 88 patients with FTAS diagnosis. As input data, on the base of which prognosis is formed, the 13 most significant risk factor, got under primary examination, were selected. Based on result of the repeated survey (2-4 year after the first appearance of the pain) patients were divided into 3 classes: favorable prognosis for a disease, including regress of the disease and 1-st functional class (FK) of stenocardia, medium prognosis - 2-nd or 3-rd FK, and unfavorable prognosis - 4-th FK, myocardial infarction or fatal outcome.
The neural network model was designed. Design, education and testing of network were realized with use software package NeuroPro. It was formed training sample, consisting of 60 patient records with the known prognosis. After learning given neural network allows to solve the problem of patients separation by quality of the prognosis with probability 100% for unfavorable prognosis, 76% for favorable prognosis and 85% for medium prognosis.
Given neural network model with sufficient degree of validity allows to form the prognosis of the evolution of FTAS. Moreover the most certain is a detection of patient with high risk of the myocardial infarction and ventricular fibrillation.

(Abstract Control Number: 317)