SA-CycleGAN-2.5D: Self-Attention CycleGAN with Tri-Planar Context for Multi-Site MRI Harmonization

2026-03-17

AI Summary

This study addresses the challenge of intensity distribution shifts in multi-center MRI caused by scanner variability, which severely compromises the reproducibility of radiomics. To mitigate this issue, the authors propose a novel CycleGAN framework that integrates 2.5D triplanar manifold injection, a global self-attention U-ResNet generator, and a spectrally normalized discriminator to achieve intensity normalization while preserving pathological structures. The method combines triplanar contextual modeling, dense self-attention mechanisms, and Lipschitz constraints, underpinned by HΔH-divergence theory for domain adaptation. Evaluated on a cohort of 654 glioma cases, the approach reduces the maximum mean discrepancy (MMD) by 99.1% and lowers domain classification accuracy to 59.7%—near random chance—demonstrating substantially improved cross-center image consistency.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Intelligent Robots: Multi-Robot Systems

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Multi-site neuroimaging analysis is fundamentally confounded by scanner-induced covariate shifts, where the marginal distribution of voxel intensities $P(\mathbf{x})$ varies non-linearly across acquisition protocols while the conditional anatomy $P(\mathbf{y}|\mathbf{x})$ remains constant. This is particularly detrimental to radiomic reproducibility, where acquisition variance often exceeds biological pathology variance. Existing statistical harmonization methods (e.g., ComBat) operate in feature space, precluding spatial downstream tasks, while standard deep learning approaches are theoretically bounded by local effective receptive fields (ERF), failing to model the global intensity correlations characteristic of field-strength bias. We propose SA-CycleGAN-2.5D, a domain adaptation framework motivated by the $HΔH$-divergence bound of Ben-David et al., integrating three architectural innovations: (1) A 2.5D tri-planar manifold injection preserving through-plane gradients $\nabla_z$ at $O(HW)$ complexity; (2) A U-ResNet generator with dense voxel-to-voxel self-attention, surpassing the $O(\sqrt{L})$ receptive field limit of CNNs to model global scanner field biases; and (3) A spectrally-normalized discriminator constraining the Lipschitz constant ($K_D \le 1$) for stable adversarial optimization. Evaluated on 654 glioma patients across two institutional domains (BraTS and UPenn-GBM), our method reduces Maximum Mean Discrepancy (MMD) by 99.1% ($1.729 \to 0.015$) and degrades domain classifier accuracy to near-chance (59.7%). Ablation confirms that global attention is statistically essential (Cohen's $d = 1.32$, $p < 0.001$) for the harder heterogeneous-to-homogeneous translation direction. By bridging 2D efficiency and 3D consistency, our framework yields voxel-level harmonized images that preserve tumor pathophysiology, enabling reproducible multi-center radiomic analysis.

Problem

• multi-site MRI harmonization
• scanner-induced covariate shift
• radiomic reproducibility
• intensity distribution shift
• neuroimaging

Innovation

• Self-Attention
• 2.5D Tri-Planar
• CycleGAN
• Global Receptive Field
• MRI Harmonization