TECHNOLOGY FOR ANALYZING ACOUSTIC DATA FOR SIGNS OF COVID-19 DISEASE

摘要

FIELD: medical diagnostics. ;SUBSTANCE: invention relates to the field of information and communication technologies (ICT) specifically designed for medical diagnostics, in particular to the diagnosis of coronavirus infection (COVID-19) based on the analysis of acoustic data of the patient using deep learning methods to identify acoustic signs caused by concomitant coronavirus infection changes in the patient’s respiratory tract. A method is proposed in which the regression problem is solved by deep learning methods, determining the probability of coronavirus infection (COVID-19) that affects the human respiratory tract using a patient’s cough, breathing and speech records. In the claimed invention, an ensemble of recurrent neural networks RNN with LSTM, the attention mechanism and linear layers, and with the convolutional neural network CNN as an encoder are used for diagnosing COVID-19, and the diagnosis is made based on decisions on three branches: cough, breathing and speech. ;EFFECT: invention provides a method for rapid diagnosis of coronavirus infection (COVID-19) in a patient with great accuracy. ;5 cl, 9 dwg