Goriacha V. Models, Methods for Information Forecasting Technology of Patients Condition in the Monitoring Medical Systems

Object of research is the processes of predicting the patients’ state in medical monitoring systems; goal of research is to improve the state prediction of patients by developing mathematical models, computational methods for predicting patients' condition and implementation them to decision support information technology in the medical monitoring; methods of research - principles of system analysis, methods of mathematical modeling (methods of artificial intelligence - artificial neural networks), approaches of information theory and methods of mathematical statistics (factor analysis), theory of formal and algorithmic systems; the result - the actual scientific task is solved, which is to develop models, methods of forecasting and information technology of decision-making support for improving the quality of the forecast in medical monitoring systems, through the creation and implementation of a computer interactive certified decision support system "RMICP®" in predicting patient status in conditions of input data parametric uncertainty , which is aimed at a wide range of healthcare professionals working on the problems of the systematic choice of informational controlled variable states and individually for each patient and recognition of his condition based on monitoring data; novelty - a method for solving the nonlinear problem of predicting the patient's state was developed for the first time basis on the concept of trend analysis, which, unlike the known ones takes into account changing the informativeness of controlled variables depending on the state (stage of the disease) of patients, which allows to assign an individual treatment program; a method of estimating the informativeness of variables of multidimensional diagnostic models and state control models was improved, which takes into account the accuracy of measuring the variables of the state and the existence of a pair correlation between them (cointegration of "partial" time series), which enables to increase the reliability of diagnostic models; a statistical classification method for patients in the medical monitoring systems was improved using the probabilistic neural networks, which includes procedures orthogonalization and dimension reduction of the state variables space and which uses Student's statistic in the selected reference basis which ensures a reduction in the probability of errors in determining the stages of the patient's condition (belonging to one of the probable states of classes) based on the monitoring data; the information technology of process automation of the of predicting patients' condition was further developed with the help of computer decision support system in the conditions of input data parametric uncertainty, which makes it possible to improve the prediction quality in medical monitoring systems. the degree of implementation - results are implemented in the educational process of the Department of Informatics of the National Aerospace University named after M.E. Zhukovsky "Kharkiv Aviation Institute", at the Department of General and Children's Oncological Urology, Medical Academy of Postgraduate Education, at the Institute of Medical Radiology named after S.P. Grigoriev; field of application - medical decision support systems.