Acquisition of interpretable domain information during brain MR image harmonization for content-based image retrieval

2025-10-16

AI Summary

To address domain shift in multi-center brain MR images caused by variations in imaging equipment and acquisition protocols, this paper proposes an interpretable domain-cooperative harmonization framework. Methodologically, we design a dual-encoder architecture (f_E, f_SE) to disentangle domain-invariant pathological features from domain-specific style representations, and introduce a linear style reconstruction mechanism to enable feature visualization and clinical interpretability. The framework is trained end-to-end via adversarial domain adaptation, reconstruction loss, and domain prediction constraints. Experiments demonstrate that our method achieves or surpasses state-of-the-art performance on image reconstruction, cross-domain disease classification, and domain identification tasks. Moreover, it enables quantitative visualization of domain-shared and domain-specific components, significantly enhancing model transparency and clinical trustworthiness.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Multi-modal Vision
• Application Domains: Humanities & Computational Social Science

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Web Mining and Content Analysis: Large pretrained models with web data
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Medical images like MR scans often show domain shifts across imaging sites due to scanner and protocol differences, which degrade machine learning performance in tasks such as disease classification. Domain harmonization is thus a critical research focus. Recent approaches encode brain images $oldsymbol{x}$ into a low-dimensional latent space $oldsymbol{z}$, then disentangle it into $oldsymbol{z_u}$ (domain-invariant) and $oldsymbol{z_d}$ (domain-specific), achieving strong results. However, these methods often lack interpretability$-$an essential requirement in medical applications$-$leaving practical issues unresolved. We propose Pseudo-Linear-Style Encoder Adversarial Domain Adaptation (PL-SE-ADA), a general framework for domain harmonization and interpretable representation learning that preserves disease-relevant information in brain MR images. PL-SE-ADA includes two encoders $f_E$ and $f_{SE}$ to extract $oldsymbol{z_u}$ and $oldsymbol{z_d}$, a decoder to reconstruct the image $f_D$, and a domain predictor $g_D$. Beyond adversarial training between the encoder and domain predictor, the model learns to reconstruct the input image $oldsymbol{x}$ by summing reconstructions from $oldsymbol{z_u}$ and $oldsymbol{z_d}$, ensuring both harmonization and informativeness. Compared to prior methods, PL-SE-ADA achieves equal or better performance in image reconstruction, disease classification, and domain recognition. It also enables visualization of both domain-independent brain features and domain-specific components, offering high interpretability across the entire framework.

Problem

• Addressing domain shifts in brain MR images across scanners
• Improving interpretability of domain harmonization for medical applications
• Preserving disease-relevant information during image harmonization process

Innovation

• Pseudo-Linear-Style Encoder Adversarial Domain Adaptation framework
• Two encoders extract domain-invariant and domain-specific features
• Reconstructs images by summing domain-invariant and specific components