Identification of the boundaries of air breathing organs on computed tomography images using convolutional neural networks of the U-NET architecture
Abstract
Identification of the boundaries of air breathing organs on computed tomography images using convolutional neural networks of the U-NET architecture
Incoming article date: 15.12.2022This article discusses the problems of determining the organs of air respiration on computed tomography images using convolutional neural networks of the U-NET architecture. The prospects of using neural networks in the analysis of medical images, as well as the use of the U-NET architecture for semantic segmentation of images are described. The structure of an artificial neural network based on the U-NET architecture is being formed. The structure of the layers of this network is visualized and the components of this structure are described. Special attention is paid to the description and implementation of the convolution process. The formula for determining the weight coefficients of the separation boundary is presented. Algorithms for the formation of an artificial neural network model and an algorithm for constructing layers are proposed. A method of increasing data for a training sample of images of medical images is considered. The image of the result of the determination of the chest organs and the corresponding mask are presented.
Keywords: convolutional neural networks, U-NET architecture, deep learning, image recognition, machine learning