Table 4.

Federated machine learning implementations in CVDs prediction.

Ref	Year	Type	Parameter Studied	Predicted outcome	Model	FL Architecture	Contribution	Dataset Used	Performance
	2018	Classification	Electronic health records	Hospitalization for CVD patients	Federated optimization scheme (cPDS) for solving sparse support vector machine		Scalability Privacy	Electronic heart records from the Boston Medical Center	Best 0.78 AUC
	2020	Regression	Heart rate	Heart rate	Federated; earning based on sequential Bayesian method (FD Seq Bayes) Empirical Bayes-based hierarchical Bayesian method (FD HBayes-EB)	Centralized Decentralized	Privacy Scalability	Private	-
	2021	Regression	Blood pressure	Blood pressure	Time-series-to-time-series generative adversarial network (T2T-GAN) (based on LSTM)	Centralized	Novelty Privacy	Cuff-Less blood pressure estimation [106] University of Queensland vital signs dataset [107]	Mean error of 2.95 mmHg and a standard deviation of 19.33 mmHg
	2021	Classification	ECG	Arrythmias	Customized alignment Model	Centralized	Personalization Privacy	Private	Accuracy: 87.85%
	2021	Classification	Electronic health records	Cardiovascular risk	Sequential pattern mining (SPM) Based Framework	Centralized Decentralized	Privacy	Nursing Electronic Learning Laboratory (NeLL)	-
	2022	Classification	ECG	Arrythmias	1D-convolutional neural Networks	Centralized	Privacy Explainability Communication cost reduction Personalization	MIT-BIH arrhythmia Database [111]	Accuracy: 98:9%
	2022	Classification	Cardiovascular magnetic resonance images	Hypertrophic cardiomyopathy	3D-convolutional neural networks	Centralized	Privacy	M&M challenge [113] ACDC challenge [114]	Best 0.89 AUC