Citation

Sen, Anik and Heng, Swee Huay and Tan, Shing Chiang (2026) Privacy-Preserving Federated Learning for MRI-Based Brain Tumour Detection Using Xception CNN. In: 2026 International Conference on Advances in Artificial Intelligence and Machine Learning, AAIML 2026, 20 March 2026 - 22 March 2026, Tokyo.

Text 20.pdf - Published Version Restricted to Repository staff only Download (1MB)

Abstract

The early detection and classification of brain tumours are essential for improving patient outcomes. Magnetic Resonance Imaging (MRI) is widely used for brain tumour diagnosis due to its high-resolution imaging capabilities. However, interpreting these images often requires significant expert input from radiologists. This study proposes a federated learning (FL) approach combined with the Xception convolutional neural network (CNN) for secure and efficient brain tumour classification from MRI scans. The integration of FL addresses privacy concerns by ensuring that sensitive patient data is not shared across institutions, while the Xception model enhances the accuracy of tumour detection. The proposed approach is evaluated on decentralised MRI datasets, overcoming the challenges of data heterogeneity and achieving an accuracy of 98% in detecting different brain tumour types, such as glioma, pituitary, and meningioma. The results demonstrate that FL with Xception offers a scalable and privacy-preserving solution for brain tumour detection in multi-institutional settings, contributing to more secure Artificial Intelligence (AI) applications in healthcare.

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Brain tumour detection, privacy-preservation, MRI Imaging
Subjects:	R Medicine > R Medicine (General) > R858-859.7 Computer applications to medicine. Medical informatics
Divisions:	Faculty of Information Science and Technology (FIST)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	01 Jul 2026 06:41
Last Modified:	01 Jul 2026 06:41
URII:	http://shdl.mmu.edu.my/id/eprint/16180

Downloads

Edit (login required)