Classification of Chronic Obstructive Pulmonary Diseases from Chest X-Ray Images Using Deep Learning

Abstract

The global Burden of Disease Study reports a prevalence of 251 million cases of COPD globally in 2016 and estimated that the Disease caused 3.17 million deaths. People in developing countries are highly at risk of the disease because they cannot get diagnosed early. In addition to this, they do not have enough experienced medical experts (radiologists). To contribute to the above severe problem, we investigated and developed an AI model that can classify sub-classes of COPD disease (asthma, Emphysema, and Chronic bronchitis). In the model development, we followed the design science methodology, which follows its scientific procedures starting from collecting the required data set to test the developed model. We have collected about 2248 images from local Hospitals, having 350 Images for each class. We have applied different image preprocessing Techniques to enhance the image. Therefore, to overcome that problem, we applied zooming, rotation, and flipping at different angles as augmentation techniques. Then Features are extracted from gray-level images using a CNN feature extraction. A classification model is built using 5 Different Pre-trained models called InceptionV3, VGG16, EffeceintNetB0, and Resnet50, including our own CNN model. The convolutional neural network architecture with the sequential model was implemented with many layers, such as convolutional, activation, and max- pooling, to extract essential features from the x-ray images. EffeceintNetB0 Accuracy was 85.7% test data, ResNet50 Model Accuracy was 67.5% on the test data, and VGG16 Accuracy was 87.6% on the test data also, our own CNN model obtained 81.1% on the test data. Experimental results show that the InceptionV3, with its filtering mechanism, has achieved a better classification performance with an accuracy of 90.1%.