Mobile Solution for Early Detection of Heart Diseases using Artificial Intelligence and Novel Digital Stethoscope

Abstract

As per the World Health Organization (WHO), cardiovascular diseases (CVD) are the number one cause of death globally, more people die annually from heart diseases than from any other cause. Moreover, in India mortality rates in rural areas is very high as compared to urban areas due to lack of proper health care facility and qualified health workers. Therefore, this paper proposes to develop a mobile, affordable system that will be used by amateur health-workers for early detection of cardiac abnormalities like Murmurs. We have designed a novel digital stethoscope to record heart sounds in real-time and upload it to our mobile application for concurrent analysis on the Cloud. To detect abnormalities, heart sounds are pre-processed via algorithms like audio slicing and segmentation, after which they will be converted into high-quality spectrograms and then classified by our pre-trained Cloud-based Convolution Neural Networks (CNN). The CNN model is trained using datasets provided by the University of Michigan, PhysioNet Challenge 2016 and PASCAL's Classifying Heart Sounds Challenge.