Federated fusion learning with attention mechanism for multi-client medical image analysis

Muhammad Irfan, Khalid Mahmood Malik, Khan Muhammad

Abstract: Highlights•A novel federated fusion learning (FFL) framework for multi-client data fusion.•FFL as a solution for statistical heterogeneity and label scarcity in clients’ data.•A multi-client latent space fusion module for various modality latent space fusion.•Detailed investigation of 19 diverse medical image datasets for generalization.•Excellent generalizability and promising performance of the global model via FFL.