Multi-Disease Detection in Chest X-Rays Using Image Processing and DenseNet-121

Abstract

One of the most popular and less costly tools of diagnosing lung related diseases is the chest X-rays. Nonetheless, the manual analysis is time consuming and it is subject to error particularly in resource constrained environments. Most of the existing models only deal with pneumonia or they deal with complex feature engineering that restricts their scalability. This project suggests implementing a hybrid solution, which is based on the combination of image processing and deep learning model, DenseNet-121, to classify the chest X- rays with multiple labels. The preprocessing such as noise reduction and contrast enhancement enhance clarity of the image which assist the model to identify various conditions such as pneumonia, Edema, Tuberculosis and fibrosis. To achieve the model transparency, it has implemented Grad CAM to generate visual explanation of the disease predictions by highlighting the important areas in the X-ray. The suggested system provides high diagnostic, better interpretability and real-time performance, which reduces the workload of radiologists and facilitates quicker clinical decisions. This method proves the useful and transformative nature of artificial intelligence in improving the diagnostics of chest X-ray and the development of support systems in healthcare.