Using Cough Sounds for the Recognition of Multiclass Respiratory Diseases With Artificial Intelligence: A Review

The multi-class identification of respiratory diseases based on cough sounds plays a critical role in the diagnosis of respiratory conditions. Compared to traditional auscultation methods, multi-class techniques integrate various deep-level disease features of cough sounds, enabling a more comprehensive capture of cough characteristics associated with specific diseases, thereby effectively reducing the risk of misdiagnosis. Furthermore, multiclass classification enhances the system’s ability to differentiate among multiple respiratory conditions, significantly improving diagnostic accuracy and robustness. This study provides an in-depth examination of artificial intelligence-driven methodologies for multi-class respiratory disease recognition. We reviewed 19 articles from a 5-year open-access literature database to summarize the current research on diagnosing voice-affecting disorders through AI techniques using cough sound samples for multi-class classification. The study also covers AI workflows for multi-class cough recognition, including datasets, feature extraction techniques, models, and evaluation metrics. Additionally, recent advancements in AI technologies, cough monitoring applications, and wearable devices are discussed, along with the challenges and future research directions aimed at developing robust and universally applicable solutions for multi-class diagnosis.