Regularization by Neural Style Transfer for MRI Field-Transfer Reconstruction with Limited Data

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Deep Learning

Background:

• Deep learning significantly advanced MRI reconstruction but requires large, specific datasets.
• Reconstruction in data-limited scenarios remains a critical challenge.
• Existing methods like regularization by denoising (RED) use denoisers as priors.

Purpose of the Study:

• To introduce Regularization by Neural Style Transfer (RNST) for MRI field-transfer reconstruction.
• To address the challenge of data-limited settings in MRI reconstruction.
• To generate high-field-quality MRI images from low-field inputs without paired training data.

Main Methods:

• Integrated a neural style transfer (NST) engine with a denoiser.
• Developed a novel framework named RNST.
• Leveraged style priors from NST to overcome data limitations.

Main Results:

• RNST successfully reconstructed high-quality MRI images across axial, coronal, and sagittal planes.
• Achieved superior image clarity, contrast, and structural fidelity compared to low-field references.
• Demonstrated robustness even with imperfect alignment between style and content images.

Conclusions:

• RNST offers a scalable and data-efficient solution for MRI field-transfer reconstruction.
• The framework shows significant potential for resource-limited clinical settings.
• RNST enables high-field-quality image generation without requiring large, task-specific datasets.