A Federated Approach Towards Detecting ECG Arrhythmia

The integration of artificial intelligence (AI) in healthcare has enabled the interest in utilizing machine learning technologies to enhance the accuracy as well as efficiency of medical diagnostics. This research study explores the application of federated learning, a decentralized machine learning paradigm combined with a ResNet model to perform electrocardiogram (ECG) arrhythmia detection. Through this approach, model training occurs locally on distributed datasets residing across multiple institutions for enhancing the model performance. Also, this study simultaneously obtains faster results and increases data security to sensitive records. This research has obtained an accuracy of 99.2% which outperforms other existing models. Finally, future research directions aimed at optimizing federated learning protocols, exploring advanced neural network architectures are outlined by extending the applicability of federated learning to other cardiac conditions and medical domains.