A Decentralized Smart Healthcare Monitoring System using Deep Federated Learning Technique for IoMT

Abstract The likelihood of privacy and security problems rises as a result. Additionally, it performs poorly due to a lack of datasets. By utilising a new Federated Learning (FL) model, this research work enhances privacy and decentralises the analysis of medical data. We created a three-tier architecture for our IoMT: In first tier, the medical data is generated by medical sensors. In first, tier we introduced data normalisation technique in order to eliminate the redundant data and uncertain data. This process reduces the complexity of the system. In the second tier, the data is submitted to the decentralised edge servers where the Deep Learning (DL) algorithm is employed. By FL model, each DL is trained with the limited data samples. The DL approach used here is Recurrent Neural Network (RNN) model. The RNN is trained to detect abnormalities in the data received from the first tier. In the final tier, the data is further stored in repositories and enabled to end users. The overall system is analysed with COVID-19 data and proved efficacy in accuracy, precision, recall and f-score.