Lungs Disease Classification using VGG-16 architecture with PCA

Abstract

People all over the world are afflicted by lung disease, which is a prevalent illness. The earliest possible diagnosis of lung illness is necessary. Due to this, a number of deep learning models for processing image data evolved over time. Advances in deep learning have helped identify lung disorders and detect them in diagnostic images. Various types of modern deep learning techniques, including vanilla neural networks, convolutional neural networks (CNN), visual geometry group (VGG) dependent neural networks, and capsule networks, can be used to classify lung cancer. The basic CNN performs poorly when trying to handle rotated, curved, or other unusual image orientations. As a result, the proposed work explored using principal components analysis (PCA) and the VGG16 deep learning architecture. In order to extract significant features from an image dataset, PCA is generally used. The Chest X-ray of National Institutes of Health (NIH) is taken as dataset which contains 112,120 images of X-ray of 30,805 different patients. In the current work, accuracy is used to evaluate performance, and VGG 16's accuracy is 79.1%. The PCA approach has raised it by up to 96%. Additionally, the proposed architecture is contrasted with current work.