Towards Accurate Auscultation Sound Classification with Convolutional Neural Network

Deep learning algorithms have been incorporated into conventional auscultation procedures, advancing the area of respiratory sound analysis. Convolutional Neural Network and Gated Recurrent Unit deep learning models were used in a study to increase the precision and effectiveness of respiratory sound categorization (GRU). The work involves developing the models using a sizable dataset of auscultation sounds and assessing how well they classified the sounds into six categories-healthy, bronchiectasis, bronchiolitis, COPD, pneumonia, and URTI-by using auscultation sounds as input. The outcomes indicated that the accuracy of the CNN model was 95%, while the accuracy of the GRU model was 93%. This effort has significantly aided in creating a reliable and effective respiratory sound categorization system.