Chapter 15 - Clinical decision support systems to improve the diagnosis and management of respiratory diseases

Abstract

The burden associated with respiratory diseases is increasing and will likely grow exponentially in the next generations. Therefore a great research effort has been directed to the development of improved methods for diagnosis and management of respiratory diseases. Advances in clinical decision support systems based on machine learning (ML) algorithms have opened a new realm of methods in this area. These methods are closely related to personalized medicine, providing support to medical decisions, practices, interventions, and technologies that are tailored to individual patients on the basis of their predicted response or risk of disease. In particular, pattern analysis of pulmonary function has attracted attention as an approach to early detection of respiratory diseases. This article presents a review of the development of clinical decision support systems using in pulmonary function analysis from a machine learning perspective. Initially, a brief description of the main ML algorithms is presented, as well as the main methods used for pulmonary function exams. Then, we discuss the previous studies concerning the use of ML methods in pulmonary function analysis in a historical order, emphasizing the state of the art in this field. Methods using respiratory data beyond pulmonary function were also reviewed, as well as recent works integrating telemonitoring of pulmonary function and ML algorithms to optimize control of disease trajectory and prevent exacerbations. This review showed that ML has been successfully used in the automated interpretation of pulmonary function tests. In several studies, the introduction of ML methods increased diagnostic accuracy, including the early diagnosis. Very promising results were also observed concerning the prevention of exacerbations. Several of these studies, however, are small-scale studies, so large-scale studies are still needed to validate current findings and to boost its adoption by the medical community. Finally, we conclude and examine important future directions for this research field, including big data analytics, interactive machine learning, and deep learning.