Publication date: 2020

DOI: 10.2991/aisr.k.201029.060

The article discusses the use of deep neural networks for analysis and recognition of images of computed lung tomography. The recognition is carried out in two stages: segmentation of lung image on the CT slice and search of pathological entity. An algorithm for segmentation of lung images in images is proposed and a model for finding pathological formations based on machine learning methods is developed. To implement the stage of recognition of pathological formations, CNN convolutional neural network is used. The developed approach ensures that areas containing pathological formations are found on sections of pictures. Images from the public LIDC/IDRI database were used to test the model. The efficiency analysis showed on the test data the accuracy of the proposed model 0.82. The software is implemented in the programming language Python 3.6 and is cross-platform. For machine learning algorithms TensorFlow 1.14, Scikit-learn 0.22.1 packages were used.

Издатель: ATLANTIS PRESS29 AVENUE LAVMIERE, PARIS, 75019, FRANCE

Тип: Article