A Novel Transformation Through Digital Twin and Federated Learning Integration: A Case Study on Cardiovascular Disease Prediction

Rabia Arkan Yurtoğlu, Havvanur Dervişoğlu, Burak Ülver, Ruşen Halepmollası, Mehmet Haklıdır

Abstract: Industry 4.0 and the Internet of Things have a pivotal role in digitalizing data and adding value to healthcare. Digital twin and federated learning are two novel technologies poised to transform many sectors and the combination of those technologies can reshape healthcare in remarkable ways. Cloud and edge technologies continue to evolve, and thus their integration with federated learning and digital twins can also be beneficial in healthcare operations, analysis, and decision-making. In this study, we presented three digital twin architectures, that is cloud-based digital twin, edge-based digital twin and federated learning-based digital twin and compared them in terms of latency, scalability, mobility, and centralizing. Also, we presented the architectures by developing a case study that aims to predict cardiac patients, continuously monitor the risks related to heart disease, and report the risks to both healthcare professionals and users in real time. We compared the predictive performance of machine learning models trained with federated learning approach against the centralized learning approach. We achieved the best results using SVM+chi2 combination with central training both on edge-based DT and cloud-based DT. Similarly, we achieved comparable outcomes utilizing PAC+chi2 with FL-based training in terms of accuracy, recall and F1-score. Also, LR+chi2+ROS combination outperformed the other classification models with precision rates of 0.92 with central traing, whereas SGDC+chi2+ROS combination outperformed the other classification models with precision rates of 0.89.

External IDs:doi:10.1007/978-3-031-62753-8_6