A Self-Organized Operational Neural Network for Heart Sound Abnormality Detection

Automated detection of heart sound abnormalities is crucial for the initial diagnosis of cardiovascular diseases (CVDs). However, the complexity of heart sound signals often hinders the extraction of their features. In this study, we introduce a heart sound classification model based on Slef-Organized Operational Neural Networks (ONN). Unlike conventional CNN-based models, ONN exhibits a closer resemblance to biological neurons, enabling it to extract target information features in highly nonlinear operations, particularly in complex environments. To enhance the classification accuracy of heart sound signals with lower signal-to-noise ratios, we propose a novel ONN-based model architecture. Experimental results confirm the efficacy of the proposed model, demonstrating superior performance compared to traditional CNN models on the PCCD database.