U-Net: Convolution Neural Network for Lung Image Segmentation and Classification in Chest X-Ray images

Main Article Content

R ROSELIN RATNARAJ
Thamilarasi V

Abstract

Medical Image analysis and diagnosis is an indispensable part of today’s unpredictable and insanitary environment. Analysis takes long procedure to find disease and it needs screening of internal organs of human body. Clinical analysis takes more time and sometimes may result in failure. Chest X-Ray images are fundamental visual mechanism for medical world. Lung chest X-Ray images need more support from Medical Diagnosis and identification. Automatic segmentation in supervised learning needs training data and unsupervised learning approaches needs labelled images in addition to shape, size and texture variation in patients Lung chest X-ray images require more assistance and support from medical radiologist. Some machine learning approaches failed to handle natural form of raw data visualisation in chest X-ray images. Deep learning models accept and process data in their own nature. Due to this nature deep learning approaches plays major role in medical image segmentation. It accesses large set of data and executes in little time and produce expected result effectively. Central impact of this paper is to deliver appropriate segmentation architecture for lung chest X-ray images. In this paper U-NET based convolution deep neural network is used for lung chest X-ray image segmentation and classification. 512 X 512 image sizes are used for training and testing. The intention of this method is to extract lung region from chest x-ray image and classify as nodule or non-nodule images and also cancerous or non-cancerous. The result indicates that the proposed method achieve high accuracy as 89.76% for segmentation and 98.40% for nodule classification and 98.79% for cancer classification.

Article Details

How to Cite
RATNARAJ, R. R., & Thamilarasi V. (2021). U-Net: Convolution Neural Network for Lung Image Segmentation and Classification in Chest X-Ray images. INFOCOMP Journal of Computer Science, 20(1). Retrieved from http://infocomp.dcc.ufla.br/index.php/infocomp/article/view/1422
Section
Machine Learning and Computational Intelligence