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

Abstract

Respiratory diseases such as asthma, pneumonia, chronic obstructive pulmonary disease (COPD), tuberculosis, and COVID-19 require timely recognition and appropriate medical evaluation. Cough is a common symptom of many respiratory conditions and contains acoustic characteristics that can provide useful information for disease recognition. This report presents a review of Artificial Intelligence (AI)-based approaches for recognizing multiclass respiratory diseases using cough sounds. The review examines cough datasets, audio preprocessing, feature extraction, and Machine Learning (ML) and Deep Learning (DL) techniques used for cough analysis. Common methods include noise reduction, cough segmentation, Mel-frequency cepstral coefficients (MFCCs), Mel-spectrograms, Support Vector Machine (SVM), Random Forest, Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM). The review also discusses applications, advantages, limitations, and evaluation methods. Findings indicate that AI-based cough analysis has potential as a non-invasive and accessible supportive screening technology. However, challenges such as limited datasets, background noise, class imbalance, recording-device variations, similar acoustic patterns, model generalization, privacy, and clinical validation remain. Future research should focus on larger and diverse datasets, explainable AI, smartphone-based systems, multimodal analysis, and clinical validation for reliable respiratory disease recognition.