Diagnose and Prognose the Syndrome of Respiratory Disorder Through Breathing Pattern Deploying DenseNet-SVM Model

Lungs are the vital organs for respiratory health, which can be infected by contaminated air and vulnerable. Due to increasing air pollution worldwide, millions of people are at risk of contracting a severe respiratory disorder. This study aims to diagnoses and prognoses these disorders’ syndrome to start recovery early for a patient. The paper proposes a DenseNet-SVM architecture to detect and classify diseases from the spectrograms extracted from breathing sounds generated by a test subject and estimate the respiratory disorders syndromes for seven types of categories: Upper Respiratory Tract Infection (URTI), Healthy, Bronchiectasis, Pneumonia, Chronic Obstructive Pulmonary Disease (COPD), Bronchiolitis, and Lower Respiratory Tract Infection (LRTI) with the corresponding Area Under Curve (AUC) of receiver operating characteristics (ROC) is 0.99, 0.82, 0.68, 0.98, 1.00, 0.67, 0.68 respectively for the unseen tested data. The study establishes a model that can detect respiratory diseases with breathing patterns and patient information with a deep learning approach.