A Deep Learning Approach for Intelligent Diagnosis of Lung Diseases

According to the World Health Organization (WHO), lung diseases contribute to millions of fatalities globally each year. Pneumonia stands out as a leading cause, claiming the lives of approximately 2.5 million individuals. Tuberculosis accounted for 1.4 million deaths in 2019 alone. Furthermore, COVID-19, the coronavirus, has resulted in over 9 million reported fatalities worldwide as of July 2024. Addressing the burden of lung diseases is crucial for public health, emphasizing the necessity of initiatives aimed at prevention and management to enhance global health outcomes. Deep learning for medical image analysis serves as a valuable resource for both academic and industrial researchers keen on delving into medical image analysis. This study explores a range of image pre-processing techniques, including Gaussian blur, total variation denoising, histogram equalization, and CLAHE, applied to a chest X-ray image dataset. These techniques aim to enhance image quality, thereby facilitating more accurate disease classification by the model. The study focuses on multiclass classification en-compassing normal, viral pneumonia, bacterial pneumonia, tuberculosis, and COVID-19 classes. To accomplish this, four deep learning models-VGG16, VGG19, ResNet50, and DenseNet121 are employed for disease detection from chest X-ray images, with their performances compared. Results indicate that VGG16 achieves the highest accuracy, with 94.04% for training and 87.80% for testing.