ON THE POSSIBILITIES OF USING THE KVAZAR PATTERN RECOGNITION SOFTWARE PACKAGE IN SOLVING MEDICINE AND HEALTHCARE PROBLEMS

Abstract

Introduction. The relevance of the application of modern methods of processing and analysis of medical data is discussed. Attention is focused on the need for a systematic approach to data analysis and the use of multivariate analysis based on pattern recognition methods. The aim of the paper is to inform medical professionals and specialists in the analysis of medical data about the possibilities of using pattern recognition methods and the KVAZAR software package in solving practical problems of diagnostics, forecasting and searching for patterns in research materials and statistical reporting data. Materials and methods. Pattern recognition methods are an effective tool for dealing with complex, poorly formalized problems and are successfully used in solving problems of classification, diagnostics, forecasting, and control. The KVAZAR software package, which was developed at the Institute of Mathematics and Mechanics, Ural Branch of the Russian Academy of Sciences, is a tool for solving pattern recognition problems. The possibilities of the package for solving the basic pattern recognition problems, in particular, learning from examples, taxonomy, and issues of controlling the quality of the processed data and filling gaps in them, are considered in detail. The materials for solving pattern recognition problems using KVAZAR were government statistics data as well as arrays of observations data obtained during special studies. In the two examples given in the paper, data on the state of health of the population, demography, and resources of health care systems of municipalities of the Sverdlovsk region for 2016–2019 were used. Results. The most interesting studies carried out with the help of KVAZAR are reported. The paper also deals with the problem of the construction and practical use of a model that classifies the municipalities of the region according to the mortality of the working-age population from acute cerebrovascular accidents (CVAs). The results modeling some management scenarios aimed at reducing mortality from CVAs are presented. An example of using KVAZAR for cluster analysis of municipalities of the region according to nine statistical indicators is presented. Conclusions: 1. The analysis performed with the use of the KVAZAR package shows that pattern recognition methods are an effective tool for solving problems of classification, diagnostics, and forecasting and can be successfully applied in data analysis in medicine and public health. 2. The results of mathematical modeling prove that an increase in the proportion of emergency hospitalizations of patients with arterial hypertension and chronic coronary heart disease helps to reduce the mortality of the working-age population from CVAs. 3. A multivariate taxonomic analysis of the municipalities based on nine indices has shown the presence of two large clusters that differ in the general mortality of the population and the level of equipment of the medical care system.