Development of an Automated Diagnostic System of Lung Pathologies in Lymphoma

Abstract

The article proposes components of a screening system based on the analysis of computed tomography images using convolutional neural networks (CNN), with several strategies for the accurate diagnosis of malignant pulmonary lymphoma nodes. The study provides an analysis of literature data on traditional methods of diagnosing lung pathologies and methods using artificial intelligence technologies. As a result of the study, the functional model and algorithm of the screening system, the DICOM image preprocessing module (Digital Imaging and Communications in Medicine) are presented. A data set for CNN training and testing has been created; the AlexNet CNN architecture has been trained and tested; a module for integrating the results of computed tomography image analysis into the metadata of a DICOM file has been presented.