Neural Network-Based Multi-class Model for Abnormal Heartbeat Audio Signal Detection

Detection of abnormal heartbeats, or arrhythmias, is crucial for early diagnosis and management of cardiac diseases. Traditional methods, such as manual auscultation and basic signal processing techniques, often fall short in accuracy and depend heavily on the expertise of the practitioner. In this paper, we introduce a novel neural network-based multi-class model designed to enhance the detection of abnormal heartbeat audio signals. The model leverages convolutional neural networks (CNNs) to automatically extract intricate features from heartbeat audio signals, thus eliminating the need for manual feature engineering. The proposed system preprocesses the raw audio signals by employing noise reduction techniques, normalization, and segmentation into short, manageable frames. These frames are then processed by several convolutional layers, which learn hierarchical representations of the audio features. The extracted features are subsequently classified into multiple classes using fully connected layers, employing a softmax function to ensure proper probability distribution over the classes. The performance of the proposed model is rigorously evaluated using a comprehensive dataset containing various types of abnormal heartbeats. The dataset is split into training, validation, and test sets to ensure unbiased performance evaluation. The model is trained using cross-entropy loss and optimized with the Adam optimizer, incorporating early stopping and regularization techniques to prevent overfitting. Our experimental results demonstrate that the proposed model significantly outperforms traditional methods, achieving a classification accuracy of 95.2%, precision of 94.8%, recall of 95.1%, and an F1-score of 94.9%. The study highlights the potential of deep learning approaches, specifically CNNs, in capturing the nuanced patterns in heartbeat audio signals, making it a valuable tool for clinicians. By enabling early and accurate detection of arrhythmias, the model can aid in timely intervention, thereby improving patient outcomes. These results indicate a substantial improvement over baseline methods, which typically achieve lower performance metrics. The study highlights the potential of deep learning approaches, specifically CNNs, in capturing the nuanced patterns in heartbeat audio signals that are critical for accurate classification. The findings suggest that the proposed neural network-based multi-class model could be a valuable tool for clinicians in the early detection of arrhythmias, thereby aiding in timely intervention and treatment.