DIST-CLIP: Arbitrary Metadata and Image Guided MRI Harmonization via Disentangled Anatomy-Contrast Representations

2025-12-08

AI Summary

Significant domain shift in MRI data—arising from variations in scanners, acquisition protocols, and imaging parameters—severely hampers the clinical generalizability of deep learning models. Existing image harmonization methods rely on target images, while text-guided approaches suffer from coarse-grained labels and limited, homogeneous datasets. Method: We propose a disentangled standardization framework that decouples anatomical content from contrast-style representations. Our approach introduces a pre-trained CLIP encoder to extract fine-grained contrast-aware features and integrates an adaptive style transfer module compatible with dual-modal guidance—either from images or DICOM metadata. A content–style separation architecture ensures strict anatomical preservation alongside high-fidelity style conversion. Contribution/Results: Trained end-to-end on real-world multi-center MRI data, our method substantially improves cross-center data comparability and model generalizability, outperforming state-of-the-art methods in both style fidelity and anatomical consistency.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Multi-modal Vision
• Natural Language Processing: Machine Translation, Multilinguality, Cross-Lingual NLP

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

Deep learning holds immense promise for transforming medical image analysis, yet its clinical generalization remains profoundly limited. A major barrier is data heterogeneity. This is particularly true in Magnetic Resonance Imaging, where scanner hardware differences, diverse acquisition protocols, and varying sequence parameters introduce substantial domain shifts that obscure underlying biological signals. Data harmonization methods aim to reduce these instrumental and acquisition variability, but existing approaches remain insufficient. When applied to imaging data, image-based harmonization approaches are often restricted by the need for target images, while existing text-guided methods rely on simplistic labels that fail to capture complex acquisition details or are typically restricted to datasets with limited variability, failing to capture the heterogeneity of real-world clinical environments. To address these limitations, we propose DIST-CLIP (Disentangled Style Transfer with CLIP Guidance), a unified framework for MRI harmonization that flexibly uses either target images or DICOM metadata for guidance. Our framework explicitly disentangles anatomical content from image contrast, with the contrast representations being extracted using pre-trained CLIP encoders. These contrast embeddings are then integrated into the anatomical content via a novel Adaptive Style Transfer module. We trained and evaluated DIST-CLIP on diverse real-world clinical datasets, and showed significant improvements in performance when compared against state-of-the-art methods in both style translation fidelity and anatomical preservation, offering a flexible solution for style transfer and standardizing MRI data. Our code and weights will be made publicly available upon publication.

Problem

• Addresses MRI data heterogeneity from scanners and protocols.
• Enables flexible harmonization using images or DICOM metadata.
• Improves style translation fidelity and anatomical preservation.

Innovation

• Disentangles anatomy from contrast using CLIP encoders
• Uses target images or DICOM metadata for flexible guidance
• Integrates contrast via novel Adaptive Style Transfer module