Federated Machine Learning for Cardiac Disease Detection Using Internet of Medical Things

The abundant availability of medical data from healthcare organizations increases the application of Machine learning (ML) techniques in medical field. Reliable and trustworthy ML models needs to be developed in order to maintain the efficacy of the healthcare providers. ML models are used extensively to make reliable disease predictions while also safeguarding the data collected through Internet of Medical Things (IoMT) devices. Federated Learning (FL) techniques are suitable for preserving IoMT data since it retains only the trained models and progresses with information from dispersed consumers. FL techniques have the caliber to drastically transform medical industry by making prompt disease diagnosis to enhance efficacy of therapy. The accuracy of these FL techniques decreases due to the massive data being exchanged between local and remote sites. In order to mitigate this issue, a novel approach FedEDFA, that combines Federated Machine learning with meta-heuristic optimization algorithm is proposed. Enhanced Dragonfly Optimization algorithm is utilized to optimally choose the relevant characteristics and employ it further for disease prediction. This approach improves the resilience of the system in insecure networking environments. The performance of the proposed FedEDFA is assessed by applying it to UCI Cleveland dataset for predicting cardiovascular disease in a secure and privacy preserving way. The accuracy of the proposed approach is 98.3% which is superior to the other existing methods taken for comparison.