Prognostication of coronary atherosclerosis for selection of tactics of management of patients with ischemic heart disease
Kotel'nikova, E.V.; Gridnev, V.I.; Dobgalevskiĭ, P.Ia.; Bespiatov, A.B.
Kardiologiia 44(3): 15-19
2004
ISSN/ISBN: 0022-9040 PMID: 15489843 Document Number: 575549
To create an artificial neural network for prognostication of coronary atherosclerosis and thus selection of patients requiring invasive interventions. Efficacy of prognostication made by a physician and by a model was compared basing on probabilistic estimates with consideration of individual level of risk. Nineteen clinical and instrumental parameters were entered into multifactorial analysis. These parameters included data from anamnesis, results of instrumental examination, a priori likelihood of ischemic heart disease, and magnitude of individual risk of cardiovascular events. The following factors were selected as significant at neural network prognostication of coronary atherosclerosis: result of exercise test, typical angina, history of myocardial infarction, ECG-signs of left ventricular hypertrophy and a priori likelihood of ischemic heart disease. Basing on these data two stage scheme of prognostication of coronary atherosclerosis was suggested. Thus artificial neural network suitable for screening and selection of patients for invasive interventions in ambulatory practice was created. Prognosis made with the use of artificial neural network was 1.5-3 times more accurate than that made by a physician.