Abstract

Cardiovascular Disease (CVD) remains a significant health concern, especially in regions like India, where its prevalence is on the rise. In response to this, we present a groundbreaking approach to CVD recognition using Convolutional Neural Networks (CNN) applied to an audio dataset containing heartbeats. Our study's primary objective was to detect CVD, with a particular focus on distinguishing between five distinct classes: normal, murmur, extra heart sound, artifact, and extrasystole. Our CNN model achieved an impressive 70% accuracy in classifying these heartbeat sounds. Importantly, for abnormal heartbeats, the model can accurately identify the type of anomaly and provide essential information on symptoms, treatments, and the urgency of seeking medical attention. In India, CVD poses a pressing health issue, with a high incidence of undiagnosed cases due to the limitations of traditional diagnostic methods, including their cost and time consumption. This study addresses the need for accessible, accurate, and rapid CVD diagnosis through an innovative approach. We employed a CNN architecture to process and classify audio heartbeat data from the Dangerous Heartbeat Dataset (DHD), implementing techniques such as data augmentation and cross-validation to optimize performance. The CNN model successfully classified heartbeat sounds into their respective categories and provided valuable insights for abnormal heartbeats, including symptom identification, treatment recommendations, and advice on seeking medical attention. This research has the potential to revolutionize CVD diagnosis by offering a non-invasive, costeffective, and time-efficient method that can be widely accessible. The model's accuracy and ability to provide symptom and treatment guidance have the potential to reduce the burden on healthcare systems, leading to earlier interventions and improved patient outcomes. Future research in this field should focus on expanding the dataset and improving model generalization. Practical applications include integrating this technology into portable devices for early screening and enhancing telemedicine capabilities for rural and underserved areas. In conclusion, this study demonstrates that CNN-based audio analysis can provide an accurate, accessible, and informative method for CVD detection, with the potential to transform the landscape of cardiovascular healthcare in India and beyond