ResNet based deep learning approach for chronic obstructive pulmonary disease prediction using lung sound analysis

Abstract

Chronic obstructive pulmonary disease (COPD) affects around 300-400 million people worldwide representing a critical healthcare challenge that requires early detection for effective intervention. This work introduces chronic lung analysis via audio signal prediction (CLASP), a novel framework achieving 97.90% accuracy in predicting COPD automatically through respiratory audio signal analysis. This method integrates advanced signal processing and deep learning architectures, comparing long short-term memory (LSTM), convolutional neural networks (CNN), and residual networks (ResNet) models for optimal performance. The ResNet architecture exhibits superior diagnostic capability with precision of 98.72%, recall of 96.86%, and 0.9937 area under the curve (AUC), as compared to existing methods by significant margins. These results establish a new benchmark for noninvasive COPD detection, thus enabling practical deployment in clinical settings thereby dramatically improving the patient outcomes by early detection and also reduce healthcare costs.