A techno-analytical insight on federated learning methodologies towards diagnosis of brain tumor

Abstract

Artificial intelligence (AI) has made rapid progress in addressing complex medical challenges, including life-threatening conditions such as brain tumors. Recent years have witnessed significant contributions from machine learning (ML) and deep learning (DL), yet practical deployment remains limited due to privacy concerns, data heterogeneity, and lack of collaborative training. Federated learning (FL) offers a promising alternative by enabling distributed training across institutions without data sharing, thereby improving detection accuracy while preserving patient privacy. This paper systematically reviews FL in the context of brain tumor diagnosis, with a focus on its mathematical foundations, core modelling approaches, and emerging research trends. The analysis highlights that while FL demonstrates strong potential in enhancing classification, detection, and segmentation tasks, major gaps remain in handling non-independent and identically distributed (non-IID) data, cross-modal integration, scalability, and real-world deployment. The key insight of this study is that future progress will rely on hybrid FL systems, fairness-aware aggregation, and security-enhanced frameworks to achieve clinically viable, equitable, and scalable diagnostic solutions.