Edge-based Heart Disease Prediction using Federated Learning

Cardiovascular diseases are one of major causes for death globally. Prediction of these diseases becomes a bit complex in the fields like clinical analysis. It is observed that over many millions of deaths are recorded because of heart disease. And it is in the ratio of four in five cardiovascular deaths are due to heart failure. In recent times for making many clinical decisions and predictions from large amount of medical data produced from healthcare industries, machine learning is being effectively used. Despite the hype, still the existing machine learning based heart disease detection methods need their data to be present in a centralized place. Since the data is from hospital, there are various privacy and security concerns that are needed to be considered and hence it becomes impossible to collect all the data and store it in one place centrally. So, with this problem statement, this research work aims to implement a federated learning approach to train a disease detection model. A shared model of federated learning that makes its averaging algorithm to perform aggregates all its local updates from its clients along with the edge device to ensures its privacy and security. The results indicate that the proposed model has achieved 93.4% of accuracy levels by integrating the LASSO feature selection algorithm.