The features of using neural networks for classification of respiratory diseases on X-ay images

Abstract

In modern medicine, automated processing and analysis of medical images is a very relevant task, as it allows to increase the accuracy of diagnostics, reduce data processing time and reduce the burden on medical personnel. This paper investigates the use of neural networks to automate the process of classifying respiratory diseases based on X-ray images. The main focus is on the development and optimization of the architecture of convolutional neural networks (CNN), which allow detecting signs of diseases with high accuracy. It is expected that the proposed system will help medical professionals in quickly and accurately detecting diseases, which will contribute to reducing mortality. In addition, the paper investigates the impact of changing the values of hyperparameters on the quality of using the MobileNetV2 neural network, pre-trained on a large ImageNet dataset, to solve the problem of multi-class classification of medical images. The use of a pre-trained model allows for the effective use of the obtained universal features and significantly accelerates the learning process on specific medical data.