Task-Adaptive 3D Cross-Field MRI Translation via Field-Conditioned Content-Style Pretraining
2026-08-10 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors address a problem in MRI scans where images look different depending on the magnetic field strength used, making it hard to compare or combine scans from different machines. They created a method that can change MRI images from one field strength to another while keeping the subject's anatomy correct, even without having exact paired examples for training. Their approach uses a 3D neural network that separates the anatomy from the scanner-specific features and learns to translate between various field strengths. Tests showed their method works well across multiple MRI types and field strengths, preserving important 3D anatomical details.
Magnetic resonance imaging (MRI)Magnetic field strengthDomain shiftCross-field translation3D neural networksAdversarial learningCycle-consistencyContent-style disentanglementAnatomical fidelityGenerative models
Authors
Haowen Pang, Yingqi Hao, Pengli Zhu
Abstract
Magnetic field strength is a major source of domain shift in magnetic resonance imaging (MRI), affecting signal-to-noise ratio, tissue contrast, spatial detail, and the visibility of anatomical boundaries. The MRIxFields 2026 challenge investigates this problem through cross-field MRI translation across acquisitions at 0.1T, 1.5T, 3T, 5T, and 7T. Its three tasks, Any-to-7T, 0.1T-to-High, and Any-to-Any synthesis, require the generation of target-field image characteristics while preserving subject-specific anatomy. This problem is particularly challenging because paired acquisitions of the same subject across multiple field strengths are rarely available for training. We propose a 3D unpaired cross-field MRI translation framework based on field-conditioned content-style pretraining. The proposed framework first learns controllable field-to-field translation across all available field strengths by disentangling anatomical content from field-dependent contrast characteristics. The pretrained backbone is then adapted to task-specific target domains. Our model comprises a 3D content encoder, a 3D style encoder, a field-conditioned style generator, an AdaIN-modulated decoder, and a multi-field discriminator. Adversarial learning encourages realistic target-field appearance, while cycle-consistency, identity, content, style, and diversity constraints promote anatomical fidelity and controllable translation. We evaluate the proposed method on MRIxFields data spanning five field strengths and three MRI modalities. Experiments on paired test data demonstrate that the framework can adapt to the three challenge settings while preserving three-dimensional anatomical structure in the synthesized volumes. The implementation code is publicly available at https://github.com/Idea89560041/3D-MRI-Field-Translation.