Detection And Analysis Of Respiratory Disease Infection In Human Body Using Radiology Images Based On Machine Learning Approaches

Abstract

Radiology imaging is the most prevalent type of examination performed in radiology departments today. The automatic categorization of illnesses can help radiologists minimize their efforts while enhancing the quality of patient care. Over the past decade, medical image analysis combined with deep neural networks (DNN) has undergone a paradigm shift, demonstrating excellent performance in numerous image classification and segmentation tasks. However, challenges such as high spatial resolution, the lack of large datasets, and the need to detect a wide range of disorders pose significant research obstacles. These challenges have driven the innovative contributions presented in this thesis. newlineOur research focuses on several key objectives. First, we aim to develop an improved radiology image dataset that includes images of pneumonia, COVID-19, and tuberculosis, which are crucial for studying the spread of respiratory infections. Second, we design a Deep Neural Network (BRDD DNN) capable of binary classification to detect the presence or absence of respiratory diseases. Third, we build an enhanced multi-class respiratory disease detection model (MRDD DNN) to understand and identify patterns of respiratory illnesses in radiology images, specifically targeting COVID-19 and pneumonia, which often exhibit similar signs and symptoms. Lastly, we propose the RESP DNN, a robust technical solution capable of detecting and differentiating between respiratory diseases such as pneumonia, COVID-19, tuberculosis, and lung infections from radiology images. newlineThe performance and comparative analysis results of these models have been tested, and the findings indicate superior performance. These advancements in automatic illness categorization in radiology imaging have the potential to significantly improve diagnostic accuracy and patient care, addressing some of the most pressing challenges in the field of medical image analysis. newline newline