Harmonizing The Power Of Deep Learning And Traditional Radiology For The Advancement Of Lung Disease Diagnosis Through Chest X-Ray Image Classification

Abstract

Lung disorders have significant implications, causing reduced lung function and complications like breathing difficulties. Early diagnosis, especially in resource-constrained settings, remains a challenge. Chest X-ray images have gained popularity for quick disease monitoring, aided by image processing and machine learning. Deep learning for lung disease detection involves image preprocessing, training, and classification. After optimizing image quality, various deep learning techniques, including CNNs, RNNs, and VGG architectures, extract relevant features from X-ray images. A unique hybrid VGG-CNN architecture is proposed, capturing fine details and broader patterns indicative of lung disorders. This approach is evaluated using open datasets like NIH Chest X-ray data. CNN features are classified using Random Forest and Support Vector Machine models, and performance is assessed using metrics like accuracy, precision, recall, and F-measure. The hybrid SVM-RF model achieves exceptional accuracy, surpassing the standard CNNVGG-SVM model by 4.35%, while the hybrid RF model outperforms the conventional CNNVGG-RF model by 4.32%. These results highlight the method's reliability in diagnosing various lung diseases. This approach's remarkable accuracy demonstrates its potential to enhance lung disease diagnosis.