SmaRT: Style-Modulated Robust Test-Time Adaptation for Cross-Domain Brain Tumor Segmentation in MRI

2025-09-22

AI Summary

To address domain shift in cross-domain MRI brain tumor segmentation—caused by scanner heterogeneity, protocol variations, and population diversity (particularly in resource-limited regions and pediatric cohorts)—this paper proposes a source-free test-time adaptation framework. Methodologically, it integrates style-aware augmentation, a dual-branch momentum network, and anatomical structure priors: style modulation enhances domain-invariant feature representation; momentum-updated pseudo-labels improve prediction stability; and anatomical consistency constraints ensure biologically plausible segmentations. Evaluated on sub-Saharan African and pediatric glioma datasets—characterized by extreme domain shifts—the method outperforms existing source-free adaptation approaches, achieving significant gains in Dice score (+3.2%) and boundary accuracy (+4.7%). Results demonstrate superior robustness and generalization under severe domain misalignment, validating its clinical applicability in low-resource and pediatric settings.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Segmentation
• Cognitive Modeling & Cognitive Systems: Adaptive Behavior

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Web Mining and Content Analysis: Large pretrained models with web data
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

Abstract

Reliable brain tumor segmentation in MRI is indispensable for treatment planning and outcome monitoring, yet models trained on curated benchmarks often fail under domain shifts arising from scanner and protocol variability as well as population heterogeneity. Such gaps are especially severe in low-resource and pediatric cohorts, where conventional test-time or source-free adaptation strategies often suffer from instability and structural inconsistency. We propose SmaRT, a style-modulated robust test-time adaptation framework that enables source-free cross-domain generalization. SmaRT integrates style-aware augmentation to mitigate appearance discrepancies, a dual-branch momentum strategy for stable pseudo-label refinement, and structural priors enforcing consistency, integrity, and connectivity. This synergy ensures both adaptation stability and anatomical fidelity under extreme domain shifts. Extensive evaluations on sub-Saharan Africa and pediatric glioma datasets show that SmaRT consistently outperforms state-of-the-art methods, with notable gains in Dice accuracy and boundary precision. Overall, SmaRT bridges the gap between algorithmic advances and equitable clinical applicability, supporting robust deployment of MRI-based neuro-oncology tools in diverse clinical environments. Our source code is available at https://github.com/baiyou1234/SmaRT.

Problem

• Addressing domain shifts in MRI brain tumor segmentation across scanners
• Improving adaptation stability for pediatric and low-resource patient cohorts
• Ensuring anatomical fidelity under extreme domain shift conditions

Innovation

• Style-aware augmentation to mitigate appearance discrepancies
• Dual-branch momentum strategy for stable pseudo-label refinement
• Structural priors enforcing consistency, integrity, and connectivity