Federated Transformer-GNN for Privacy-Preserving Brain Tumor Localization with Modality-Level Explainability

TL;DR

This paper introduces a federated Transformer-GNN framework for privacy-preserving brain tumor localization that leverages multi-institutional data, enhances model performance, and provides modality-level explainability aligned with clinical insights.

Contribution

It presents a novel federated learning approach combining Transformer and GNN architectures for brain tumor localization with explainability analysis in healthcare.

Findings

Federated learning matches centralized performance in tumor localization.

Isolated training triggers early stopping, limiting model improvement.

Deeper network layers focus more on T2 and FLAIR modalities.

Abstract

Deep learning models for brain tumor analysis require large and diverse datasets that are often siloed across healthcare institutions due to privacy regulations. We present a federated learning framework for brain tumor localization that enables multi-institutional collaboration without sharing sensitive patient data. Our method extends a hybrid Transformer-Graph Neural Network architecture derived from prior decoder-free supervoxel GNNs and is deployed within CAFEIN\textsuperscript{\textregistered}, CERN's federated learning platform designed for healthcare environments. We provide an explainability analysis through Transformer attention mechanisms that reveals which MRI modalities drive the model predictions. Experiments on the BraTS dataset demonstrate a key finding: while isolated training on individual client data triggers early stopping well before reaching full training capacity, federated learning enables continued model improvement by leveraging distributed data, ultimately matching centralized performance. This result provides strong justification for federated learning when dealing with complex tasks and high-dimensional input data, as aggregating knowledge from multiple institutions significantly benefits the learning process. Our explainability analysis, validated through rigorous statistical testing on the full test set (paired t-tests with Bonferroni correction), reveals that deeper network layers significantly increase attention to T2 and FLAIR modalities ($p<0.001$, Cohen's $d$=1.50), aligning with clinical practice.