Abstract

Respiratory sound auscultation is an important component of pulmonary disease diagnosis, yet accurate interpretation requires significant clinical experience and can be
difficult for medical trainees. This paper presents an intelligent educational platform designed to support respiratory sound learning through integrated audio analysis, visualisation, and automated feedback. The system enables recording, storage, and replay of lung sounds whileproviding synchronised waveform and spectrogram visualisations to support auditory–visual learning. A convolutional neural network (CNN) with a ResNet backbone is employed to identify abnormal respiratory sound events. In addition, a lightweight CNN-based model is used to perform respiratory disease classification from the same recordings. The analysis results are presented through an interactive dashboard that supports guided listening, segment review, and structured examination summaries