Abstract
Introduction: The outbreak of the coronavirus infection, which has escalated into a pandemic, has worsened the already unfavourable situation with respiratory system diseases in Ukraine. The burden on doctors has significantly increased, necessitating the exploration of simplified and expedited methods for conducting routine respiratory examinations. The research aims to describe a model for creating an automated differential diagnosis of respiratory noise using an electronic stethoscope, combining medical and clinical information about the types of respiratory noise characterizing the normal or pathological state of the respiratory system with a means of its information and technical processing.
Material and methods: The research methods were analysis of theoretical information about the types of respiratory noise, analysis of technical information for choosing an information technology tool for processing biological signals; synthesis of the results; modelling.
Results: The research resulted in a model of automated differential diagnosis based on the principle of auscultation, which includes the process of extracting the sound of air movement inside and outside the lungs and the classification of the extracted sounds. Automation of this process concerned only the classification of the extracted sounds since the principle of extraction itself was the same for both mechanical and automatic implementations.
Conclusions: The automatic classification process was intended to reduce the time of the procedure and reduce the influence of the human factor, eliminating the possibility of medical error. To implement the process, a deep machine learning method was used, the array of information for which was to be a created phonotheque of acoustic signals of the respiratory system, which would include all types of respiratory noise concerning normal or pathological processes in the body.