AI-Doctor - A Machine Learning Aided Health Monitoring system by analyzing sound signals

According to the WHO, cardiovascular diseases (CVDs) cause the most fatalities. We want to establish a virtual medical diagnosis system that doctors can accomplish in places devoid of medical services using a digital stethoscope connected to the cloud (and a mobile device). In this research, we present a system called AI-Doctor that uses machine learning (ML) algorithms to analyze heart and adjacent organs sound signals to make a virtual preliminary diagnosis. AI-Doctor uses a digital stethoscope to capture and analyze sound signals in real time with a deployed trained model and responds through a mobile/web application and capture the feedback for re-tuning. The whole training pipeline can be deployed in any cloud platform (Azure services in this case) dealing with sound signals features extraction, imbalance data handling and bag of ML models. We found that Support Vector machines with Radial Basis Function kernel (SVM-RBF) identified testing samples with 74.36% accuracy and 84% average precision followed by Logistic Regression (LR) with 73.5% accuracy and 80% average precision.