Deep Residual U-Net Based Lung Image Segmentation for Lung Disease Detection

Abstract

The World Health Organization (WHO) estimated that by the year 2030, lung disorders such as Chronic Obstructive Pulmonary Disease (COPD) would be one of the leading cause of death all over the world. Consequently, accurate and timely detection of lung diseases may prevent further death. It is therefore vital that the early detection may lead to treatment and prevention of mortality among patients. However, there are only a minimum number of experts or well-trained radiologists reading Chest X-Ray (CXR) that delays the timely diagnosis of lung diseases. In order to aid the radiologist in reading CXR images, a computer-aided tool is proposed. Before the processing of images, it needs to be segmented to make it easier for the machine to understand. This study is focused on developing a model that will segment the lung from CXR images. Using Residual U-Net (ResUnet) architecture based semantic segmentation, the researchers were able to develop and train a model using a set of 562 CXR images and lung mask images, 70% of the images were used as training data and 30% as test data. The model was trained with 40 epochs and a batch size of 16. Dice coefficient was used to assess the similarity of the segmented result and the ground truth mask. The developed model has achieved a Dice coefficient of 0.9860. The developed model can then be used in classifying lung diseases by focusing on the segmented image rather than focusing on the entire CXR image.