Improving Augmented X-ray Images for Chest Disease Diagnosis Using Deep Learning

Abstract

The diagnosis of respiratory conditions, including coronavirus disease (COVID-19) and other chest diseases, presents significant challenges due to the complex interpretation requirements of clinical data and radiological imaging. This study introduces an innovative deep learning framework for enhancing the diagnostic accuracy of chest diseases through advanced X-ray image analysis, with a particular focus on COVID-19 detection. We developed a multi-stage diagnostic model incorporating advanced image preprocessing techniques and a modified Very Deep Convolutional Networks (VGG16) architecture. Our integrated approach combines deep feature extraction, Linear Discriminant Analysis (LDA) for dimensionality reduction, and classification through Support Vector Machine (SVM) and Neural Network (NN) algorithms. The primary contribution is a novel hybrid architecture that effectively combines these components while introducing an enhanced preprocessing pipeline for improved image quality. Experimental results demonstrate the framework's effectiveness, achieving 96.3% accuracy, 95.8% sensitivity, and 96.7% specificity using neural network classification. The VGG16-LDA-SVM pipeline achieved 84.2% accuracy, representing a 12.5% improvement over conventional approaches. For COVID-19 detection specifically, the model achieved 94.5% precision and 93.8% recall. These results indicate that our proposed methodology significantly enhances diagnostic capabilities in clinical settings, offering a valuable tool for healthcare practitioners in their diagnostic decision-making processes.