FPECGNET: A deep learning framework based on federated learning and prototype learning for interpretable ECG...

Under everyday operational conditions, data privacy regulations prohibit sharing information between hospitals or institutions, contributing to the emergence of data silos. This scarcity of data hinders improvements in model performance. Another major obstacle to AI applications in medicine is the lack of interpretability, where the decision-making process of models cannot be reflected. We propose a deep learning framework based on federated learning and prototype learning that simulates the reality of hospital data silos while endowing the model with interpretability. Using the publicly available PTB-XL dataset, we divided it into three subsets by category and trained the model using the federated learning framework, achieving superior performance on these subsets. When tested on the aggregated model using the PTB-XL dataset, it demonstrated performance comparable to the current state-of-the-art model.