An Effective Lung Sound Classification System for Respiratory Disease Diagnosis Using DenseNet CNN Model with Sound Pre-processing Engine

Lung sound auscultation is a simple, inexpensive, and non-invasive method of diagnosing respiratory diseases. But the experience of each physician may be different, resulting in inconsistent diagnostic results. To solve this problem, we built a deep learning model for classifying lung sounds, which can provide physicians with a more consistent reference for accurate diagnosis. Based on lung sound dataset obtained on children aged from 1 month to 18 years old, we proposed a classification system with optimized pre-processing methods combined with a DenseNet169 CNN model. Four different classification tasks results are provided with respect to a total score given rule, 89.0% for task 1.1, 90.9% for task 1.2, 83.8% for task 2.1 and 67.3% for task 2.2.