Author:
Ilyasova N Yu,Shirokanev A S,Demin N S
Abstract
Abstract
This paper proposes a technology for lung image segmentation using textural features. To analyze the informativeness of the features, the K-Means method was used. The segmentation result can be used to describe the characteristics of the patient: age, gender, physique, etc. The study was conducted on a large set of fragmented images. X-ray clustering errors for the 12×12, 24×24, and 36×36 fragmentation windows were presented. Pre-processing of images was also used to estimate the quality. The study showed that the technology provided key objects selecting error at little more than 25%. However, the alignment procedure reduced this error to 13%. When using the k-means method, a link was found between the segmentation error and the size of the fragmentation window. The selection of the range of interest on the lungs x-ray in accordance with the equalization was considered preferable. The high error of the first clustering method was associated with the choice of feature space; in our case, image processing was conducted for all features calculated using MaZda software.
Subject
General Physics and Astronomy
Reference17 articles.
1. Lung mass density analysis using deep neural network and lung ultrasound surface wave elastography;Zhou,2018
2. Deep learning for classifying fibrotic lung disease on high-resolution computed tomography: a case-cohort study;Simon;The Lancet Respiratory Medicine,2018
3. The role of high-resolution computed tomography in the work-up of interstitial lung disease;Verschakelen;Curr Opin Pulm Med,2010
4. An integrated index for the identification of diabetic retinopathy stages using texture parameters;Acharya,2012
5. Using texture analyses of contrast enhanced CT to assess hepatic fibrosis;HeiShun,2016
Cited by
2 articles.
订阅此论文施引文献
订阅此论文施引文献，注册后可以免费订阅5篇论文的施引文献，订阅后可以查看论文全部施引文献