Multilayer Neuro-Fuzzy Network for Monitoring the Severity of Community-Acquired Pneumonia in a Telemedicine System

A neuro-fuzzy classifier of the severity of out-of-hospital pneumonia based on the CRB-65 pneumonia risk scale was synthesized. The base of fuzzy decisive rules of the fuzzy logic inference system was formed and the belonging functions for input and output variables in the selected segment of risk factors were determined. Approbation of the neuro-fuzzy model of the hybrid classifier of the degree of severity of out-of-hospital pneumonia was carried out on the experimental group of 200 patients with out-of-hospital pneumonia of various degrees of severity. The classifier model on the control sample demonstrated diagnostic sensitivity of 90%; diagnostic specificity of 86%. The results of the obtained model of the risk of out-of-hospital pneumonia were compared with the results of expert evaluation and the results obtained using known regression models. The classification quality indicators of the synthesized neuro-fuzzy classifier allow to recommend it for telecommunication systems of remote monitoring of the degree of severity of out-of-hospital pneumonia.