Automated Lightweight Model for Asthma Detection Using Respiratory and Cough Sound Signals.

Area of Science:

• Artificial Intelligence
• Medical Diagnostics
• Respiratory Medicine

Background:

• Chronic respiratory diseases like asthma and COPD present significant global health challenges.
• Accurate and timely diagnosis is crucial for effective management and treatment.
• Current diagnostic methods can be invasive or inaccessible in certain settings.

Purpose of the Study:

• To develop and validate an AI-based system for classifying asthma, COPD, and healthy individuals using cough and respiratory sounds.
• To leverage dual acoustic signals for enhanced diagnostic accuracy.
• To create lightweight, automated deep learning models for improved early detection and clinical management.

Main Methods:

• Utilized machine learning algorithms (RF, SVM, DT, NN, KNN) integrated into AI models.
• Employed 5-fold cross-validation for robust model training and validation.
• Applied Gabor time-frequency transformation for feature extraction and NCA for feature selection.
• Implemented a majority voting ensemble technique for improved diagnostic reliability.
• Conducted separate experiments using respiratory and cough sound subsets for targeted assessment.

Main Results:

• The ensemble model achieved 94.05% accuracy using respiratory sounds and 83.31% accuracy using cough sounds for asthma vs. normal differentiation.
• Demonstrated significant benefits from integrating multiple classifier models and sound modalities.
• Showcased unprecedented accuracy and robustness for future respiratory disease diagnostics.

Conclusions:

• The AI system integrating cough and respiratory sounds sets a new benchmark for respiratory disease detection.
• The dual-sound analysis approach offers advancements in early detection and management of asthma and COPD.
• The proposed model shows strong potential to transform asthma diagnostic practices and support clinical decision-making.