HyBloFED: A Hybrid Blockchain Integrated Federated Learning Approach for Brain Tumor Classification

Abstract Brain tumors, complex and potentially devastating, demand precise classification for effective patient prognosis and treatment planning. This paper introduces a novel approach to automate brain tumor classification using deep learning techniques, particularly convolutional neural networks (CNNs). However, conventional centralized methods compromise patient privacy and data security. To address this issue, federated learning (FL), a collaborative paradigm enabling model training across multiple institutions and aggregating models at a central server while preserving the confidentiality of sensitive medical data, is proposed. Moreover, an aggregation function at the central server is modified to identify the effect of aggregation on global model training. In addition to that, Blockchain technology is also integrated with FL architecture to enhance privacy preservation and trust, to ensure the integrity and immutability of patient data. By synergistically integrating modified FL, CNNs, and Blockchain technology, the proposed approach achieves accuracy (98%) and security in brain tumor classification. Through this, it aims to advance the field of medical imaging while prioritizing patient privacy and data security (through Blockchain technology) in brain tumor diagnosis and treatment.