Orientation-Aware Diffusion Super-Resolution for 3T-Like Fetal MRI from Routine 1.5T Scans

Area of Science:

• Medical Imaging
• Neuroimaging
• Biomedical Engineering

Background:

• Fetal MRI is vital for assessing early brain development.
• 3T MRI offers superior detail but faces motion and artifact challenges in fetal imaging.
• 1.5T MRI provides motion tolerance but yields lower image quality, impacting morphometric analysis.

Purpose of the Study:

• To develop a method for synthesizing 3T-like fetal brain MRI contrast from 1.5T scans.
• To bridge the quality gap between 1.5T and 3T fetal MRI without compromising motion robustness.
• To enable reliable morphometric analysis in routine clinical fetal MRI acquisitions.

Main Methods:

• An orientation-aware diffusion super-resolution framework utilizing a Swin-UNet backbone.
• Integration of gated FiLM-based orientation embeddings and a residual error-shifting diffusion mechanism.
• Training on the FaBiAN phantom with controlled image degradations for robust generalization.

Main Results:

• The framework significantly sharpens gyri and reduces partial-volume effects in both synthesized and clinical fetal MRI data.
• Improved tissue segmentation accuracy was observed using Fetal-SynthSeg after NeSVoR reconstruction.
• The proposed method consistently outperformed state-of-the-art restoration baselines.

Conclusions:

• The developed framework successfully synthesizes high-quality fetal brain MRI from lower-field 1.5T scans.
• This approach enhances the reliability of morphometric estimates for fetal brain analysis.
• It offers a pathway to achieve 3T-level diagnostic quality in routine 1.5T clinical settings.