Hybrid Neuro-Fuzzy Classifier for Monitoring the Effectiveness of Treatment of Diseases of the Respiratory System, Taking into Account Comorbidity

Abstract

The purpose of research is to develop a hybrid neuro-fuzzy classifier for remote monitoring of the severity of community-acquired pneumonia, taking into account the risk of concomitant diseases.Methods. To assess the severity of community-acquired pneumonia and determine the effectiveness of its treatment plan, a hybrid neural network is included in the hybrid neuro-fuzzy classifier, which contains three macrolayers: PNNFNN-FNN*. The number of decisive blocks of the PNN macrolayer is equal to the number of segments allocated in the space of informative features, and the output of each PNN block produces risk and non-risk assessments of communityacquired pneumonia by severity clusters. Aggregation of decisions made over N segments of the space of informative features is carried out in the FNN layer, which has the structure of a fuzzy decision-making module. The aggregation of 2L PNN-FNN outputs occurs in the FNN* macrolayer. The same macrolayer takes into account the influence of comorbidity on the severity of community-acquired pneumonia.Results. The testing of a hybrid neuro-fuzzy classifier of the severity of community-acquired pneumonia was carried out on an experimental group of patients with community-acquired pneumonia with comorbidity in the form of coronary heart disease. Indicators of the quality of classification of the severity of pneumonia taking into account the risk of comorbid disease using the example of coronary heart disease showed that the aggregation of the classifier of the severity of community-acquired pneumonia and the classifier of the risk of comorbid disease in the form of a hybrid neuro-fuzzy classifier makes it possible to improve the quality of assessing the severity of community-acquired pneumonia by more than 10% for all quality indicators.Conclusion. A hybrid neuro-fuzzy classifier, built on different pattern recognition paradigms, makes it possible to identify clusters of disease severity and improve the quality indicators for classifying the severity of community-acquired pneumonia in the presence of comorbidity by an average of 12%.