AI stethoscope for heart murmur detection and classification

Abstract

Cardiovascular diseases remain a leading cause of mortality in Bangladesh, exacerbated by a critical shortage of pediatric cardiac specialists in rural regions. This project presents the design and implementation of a low-cost, Cloud-Connected AI Stethoscope aimed at democratizing cardiac screening. The system integrates a dual-microphone setup with Active Noise Cancellation (LMS algorithm) to capture high-fidelity heart sounds, achieving a Signal-to-Noise Ratio (SNR) improvement of +12.6 dB even in noisy clinical environments. Captured audio is digitized by an ESP32 microcontroller and transmitted via Wi-Fi to a cloud server, where a Fusion Convolutional Neural Network (CNN) detects and classifies heart murmurs with 91% accuracy. By offloading computation to the cloud, the device maintains a low unit cost of approximately 5,650 BDT while ensuring diagnostic reliability. This solution offers a scalable, affordable tool for frontline health workers to identify cardiac risks early, potentially reducing preventable deaths in underserved communities.