Enhanced Privacy-Preserving Architecture for Fundus Disease Diagnosis with Federated Learning

Learning (ML) has grown exponentially. However, due to the many privacy regulations regarding personal data, sharing data from multiple sources and storing it in a single (centralized) location for traditional ML model training is often infeasible. Federated Learning (FL), a collaborative learning paradigm, can sidestep this major pill by allowing the creation of a global ML model that is trained by aggregating model weights from individual models that are separately trained on their own data silos, therefore avoiding any data privacy concerns. This study addresses the centralized data issue by applying a novel DataWeightedFed approach for effective fundus disease diagnosis from ophthalmic images. It includes a novel method for aggregating model weights by comparing the size of each model’s data and taking a dynamically weighted average of all the model’s weights. Experimental results showed a slight average of 1.85% loss in accuracy when training using FL compared to centralized ML model systems. The obtained results demonstrate that FL can achieve maximum privacy for ML in fundus disease diagnosis while only compromising a minuscule amount of accuracy, allowing for secure, collaborative ML model training within the eye healthcare space.