Using Deep Learning for pneumonia Recognition from Chest X-Ray Images

Abstract

At the moment, pneumonia is one of the deadliest and formidable illnesses, with the manifestation of severe suffering and posing a rather significant public health problem all around the world. Prompt and accurate diagnosis is the most important factor in reducing mortality rates of pneumonia because it’s symptom’s can quickly turn without being noticed. Hence that time of treatment and management is very important. Chest X-ray, as a rule, is the first step in pneumonia diagnosis, because it gives the doctor a chance to see the structure of the lungs on the chest X-ray and to identify specific signs of pneumonia like consolidations, infiltrates, or pleural effusions. Nevertheless, the images' interpretations are sophisticated and tedious , mostly requiring expert medical radiological knowledge. In order to tackle these problems, we focus on designing an advanced machine vision algorithm that can recognize and examine cases of pneumonia when viewing a chest X-ray image. Abiding by the CNN and HOG advanced techniques that are needed to automate the pneumonia identification process by acquiring and processing the relevant image features. The model is trained using an extensive amount of data from the CNN and it is integrated to the VGGNet classifier model to ensure high accuracy in detecting pneumonia regions from normal lung tissue. Thus, the algorithm will facilitate a rapid diagnosis as well as timely intervention. Additionally, in order to improve the quality of image analysis, we will use the Modified Anisotropic Diffusion Filtering (MADF) method which enables us to keep the edges and remove the noise correctly thus, paving the way for the identification of pathological features. Furthermore, a watershed algorithm will be used as part of the image processing to isolate and demarcate the large cracks or other abnormalities on the original X-ray images so that the physicians have more information on the spread and intensity of the lung damage due to the pneumonia. Through the combination of cutting-edge machine vision techniques with the conventional diagnostic approaches, our innovative tool will be capable of transforming pneumonia diagnosis to an extent that it will contribute to better out-turns and fight against the formidable ailment.