ЗАСТОСУВАННЯ ЗГОРТКОВИХ НЕЙРОННИХ МЕРЕЖ ДЛЯ ДІАГНОСТИКИ COVID-19 НА ОСНОВІ РЕНТГЕНОГРАМ ЛЕГЕНЬ

Abstract

The object of the study is the process of classification of lung radiographs for the diagnosis of COVID-19. The research is based on the use of deep convolutional neural networks, which make it possible to store spatial information and analyze complex images, preventing the attenuation of the gradient. The principle of operation of convolutional neural networks and the advantages of their use in application to complex images, in comparison with artificial neural networks based on a multilayer perceptron are considered. The main assumption of the study is the hypothesis that the use of a deep convolutional neural network for the classification of radiographs of the lungs will obtain a high-accuracy result in the diagnosis of COVID-19 and will automate the diagnostic process. The urgency of the problem of automated diagnosis of COVID-19 on the basis of lung radiographs is considered. Training of high-performance architectures of deep convolutional neural networks, with the use of additional methods of image processing to prevent retraining. The results of neural networks are compared and statistical information is given to assess the quality of their work. The analysis of the results of the artificial neural network, using image division by the Lyme method. The expediency and prospects of using deep convolutional artificial neural networks for automation of COVID-19 diagnosis on the basis of pulmonary radiographs are substantiated. Common errors of artificial neural networks and possible approaches to their prevention are analyzed. The disadvantages of using the considered approaches and the difficulties that may arise in automation are considered, according to the results, possible options for improving the quality of the deep convolutional neural network are proposed.