Previously submitted to: JMIR AI (no longer under consideration since Mar 18, 2026)
Date Submitted: Jul 16, 2025
Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
Federated Learning for Privacy-Preserving Brain Tumour Detection Using MRI and Convolutional Neural Networks
ABSTRACT
Background:
Artificial intelligence (AI) integration in medical diagnostics has shown promising results, particularly with deep learning (DL) models applied to medical imaging. However, centralized training methods require transferring sensitive patient data, posing significant privacy and regulatory concerns.
Objective:
This study proposes a federated learning (FL) framework using convolutional neural networks (CNNs) to detect brain tumours in MRI images in a privacy-preserving manner. The goal is to maintain diagnostic accuracy while ensuring data remains local to institutional nodes, in compliance with ethical and legal standards such as HIPAA and GDPR.
Methods:
MRI data from the publicly available Brain MRI Images for Brain Tumour Detection dataset was preprocessed (grayscale conversion, resizing to 64×64 pixels, normalization) and distributed across five simulated clients using the Flower FL framework within Google Colab. Each client trained a lightweight CNN locally; model parameters were aggregated using the FedAvg algorithm over five communication rounds. Performance was assessed using AUC, accuracy, precision, recall, F1-score, confusion matrices, and ROC curves.
Results:
The federated CNN achieved an AUC of 0.71 and an accuracy of 65%, with a training loss reduction from 0.52 to 0.31. Although performance was slightly lower than a centralized counterpart (AUC 0.76), the federated model effectively balanced privacy and diagnostic performance. Visual analysis confirmed meaningful differences in intensity features across classes. Data and training were retained locally, ensuring compliance with privacy protocols.
Conclusions:
The proposed FL framework demonstrates the feasibility of decentralized, privacy-preserving deep learning for medical imaging. While the current study simulates idealized client conditions, future work will incorporate real-world heterogeneity, client dropout, and explainability tools like Grad-CAM to enhance clinical applicability and interpretability.
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.