AI-Driven Innovations in Predicting Respiratory Diseases: Review of Progress and Applications

Chronic Obstructive Pulmonary Disease (COPD), a specific lung disease, causes approximately 3 million deaths per year globally and currently affects around 480 million people worldwide. However, the current method of diagnosing COPD through physical medical professional visits and checkups is both time-consuming and costly due to the need for specialized medical personnel and equipment. While application of artificial intelligence (AI) shows promise in a future direction for clinical diagnosis, machine learning model development presents significant challenges and remains in the research phase. This paper reviews progress made in the past five years in predictive AI for respiratory illnesses, especially through classical machine learning (ML) and deep learning (DL) model development. The paper first introduces seven databases used to develop predictive ML models. Secondly, it briefly compares architectures between classical ML and DL. Thirdly, the paper summarizes common ML models for prediction and provides a summary and comparison of advantages for classical ML and DL real-life applications. Finally, this paper addresses the challenges and opportunities of using AI for clinical predictions.