Med Sound AI: A CNN-LSTM Framework for Accurate Respiratory Disease Classification from Lung Sounds

Abstract

Respiratory diseases, including pneumonia and asthma together with bronchitis, create a significant worldwide health issue that particularly affects areas with limited access to trained healthcare professionals. Medical practitioners depend on their clinical skills to assess patients through lung auscultation which results in medical errors and procrastinated diagnosis. The MedSound AI project develops an automated intelligent system that detects respiratory diseases by analyzing lung sound recordings with high accuracy and early detection capabilities. The system uses a hybrid deep learning architecture which combines Convolutional Neural Networks (CNNs) for audio spectral feature extraction with Long Short-Term Memory (LSTM) networks that analyze respiratory cycle time-based patterns. The model uses publicly accessible lung sound recordings to achieve 94 percent accuracy when identifying different respiratory conditions. The proposed approach reduces the need for manual feature engineering and enhances the robustness of audio-based analysis. The trained model has been integrated into a user-friendly interface which enables both clinicians and non-specialists to upload lung sound recordings and obtain instant classification results. The framework provides three main deployment options which include accessible and portable and scalable solutions that meet requirements for both clinical settings and resource-limited environments. MedSound AI uses deep learning together with cost-effective digital solutions to help diagnosis occur quickly while bettering patient outcomes and establishing fair healthcare access for all.