A Comprehensive Review of Lung Disease Detection Using Machine Learning and Deep Learning Techniques

Abstract

Respiratory diseases, such as pneumonia, tuberculosis (TB), chronic obstructive pulmonary disease (COPD) and COVID-19, are among the world's major causes of illness and death. Timely and precise diagnosis plays a pivotal role in patient outcomes; nevertheless, conventional diagnostic processes are limited by subjectivity, resource availability, and the requirement for skilled professionals. The emergence of integrating machine learning (ML) and deep learning (DL) in the diagnosis of lung diseases has introduced a new paradigm that may improve diagnostic accuracy, shorten the diagnostic process, and broaden access to health care, especially to those in remote settings. This review paper provides an overview of the research using ML and DL techniques for the diagnosis of lung diseases, focusing on chest X-rays and computed tomography (CT) scans. This review examines popular datasets, feature extraction techniques, classification models, convolutional neural network (CNN) models, and transfer learning methods. It also explores image preprocessing techniques, data augmentation methods, and performance metrics. The research highlights the challenges in this field - lack of labeled data, class-imbalance, and interpretability of the models - and suggests future research opportunities. The work is primarily an analysis of the literature, and it does not include any original experiments or clinical studies.