Pattern recognition in evaluation of haemorheological and haemodynamical measurements in the cardiological diagnostics.

Abstract

The non-invasive differential diagnosis of ischaemic heart disease (IHD) and acute myocarditis or secondary cardiomyopathy following myocarditis can be difficult on the basis of the complaints, resting and exercise ECG and nuclear cardiological tests. 92 patients (mean age: 46 years) in the first step and 100 patients (mean age: 44 years) in the second step all with heart troubles, were examined. Besides determination of the routine parameters, nuclear haemodynamical and haemorheological measurements were carried out. Then each group of the patients was classified into 4 subgroups: 1) myocardial infarction /n:9/, 2) IHD /52/, 3) myocarditis /28/, 4) chronic cor pulmonale (CCP) /3/ subgroups in the first group and 1) normal /n:20/, 2) IHD /50/, 3) myocarditis /16/, 4) chronic cor pulmonale /14/ subgroups in the second group. The patients were reclassified by our multivariate pattern recognition algorithm (PRIMA). The average effectiveness of our method was over 80%, the recognition abilities for the subgroups (classes) ranged between 71 and 100%. An analysis of the discrimination power of the properties has made it evident that the haemorheological features were more characteristic than the haemodynamic ones in distinguishing the two differential-diagnostically critical groups. Our results show that our multivariate statistical method can be useful for the computer-aided decision in cardiological diagnostics.