Early Prediction of Chronic Obstructive Pulmonary Disease: A Deep Transfer Learning Approach

Chronic Obstructive Pulmonary Disease (COPD) is a prevalent, crippling lung ailment that often progresses undetected until its advanced stages, leading to significant morbidity and mortality. Early prediction and interference are crucial to improving patient outcomes. Here, a novel approach for the early prediction of COPD using deep transfer learning is proposed. Three cutting edge models from deep transfer learning, namely Inception, ResNet, and VGGNet, are trained to extract and classify features from audio recordings of breathing sounds. These models are fine-tuned on a dataset consisting of sounds from both healthy individuals and those with COPD, along with six other lung disorders. The extracted features are used to predict the presence of COPD, enabling early detection and intervention. The outcome of the proposed methodology demonstrates the efficiency of the model’s capability to predict COPD, with the deep transfer learning models achieving high precision in distinguishing COPD patients from non-COPD individuals. This approach shows the potential for cost-effective and early prediction of COPD, facilitating timely interventions and improved management of the disease.