An Artificial Intelligence based mobile solution for early detection of valvular heart diseases

Heart disease is the number one cause of death globally, with over three-quarters occurring in developing countries (WHO). In developing countries, majority of primary-care is handled by unqualified practitioners owing to an acute shortage of doctors. The goal of this project is to develop a portable and low-cost system that can be used by untrained frontline health-workers for early detection of valvular heart abnormalities such as arrhythmias and murmurs. Using Artificial Intelligence (AI), Cloud and Mobile, we simulated stethoscope-based cardiac auscultation typically done by trained doctors. We detect abnormal heartbeat patterns by applying pre-processing algorithms such as slicing and FFT to stethoscope sounds, converting into high-quality spectrograms and classifying with state-of-the-art Cloud-based Convolutional Neural Network (CNN) models trained with normal and abnormal heart sounds. The system was trained and tested with heart sound datasets from UMichigan, Pascal and PhysioNet and 95% average classification accuracy was achieved. We optimized spectrogram parameters with experimentation for higher accuracy. The infrastructure can be extended to other diseases such as pulmonary disorders that can be detected via auscultation. The end-to-end system was successfully cross-validated with heart sounds collected in real-life with a low-cost digital stethoscope developed as part of the project, proving that it can be deployed as a holistic diagnostic tool to improve healthcare in developing countries and beyond.