Improving Clinical Utility of Fetal Cine CMR Using Deep Learning Super-Resolution

Area of Science:

• Cardiovascular Imaging
• Medical Artificial Intelligence
• Fetal Medicine

Background:

• Fetal cardiovascular magnetic resonance (CMR) is crucial for prenatal diagnosis of congenital heart disease (CHD).
• Clinical application is limited by long scan times and fetal motion artifacts.
• Novel techniques are needed to enhance the utility of fetal CMR.

Purpose of the Study:

• To evaluate the clinical utility of deep learning (DL) super-resolution (SR) for rapid, low-resolution fetal CMR.
• To assess if DL-SR can maintain or improve diagnostic image quality and accuracy.

Main Methods:

• Prospective study of 42 fetuses with CHD undergoing third-trimester CMR.
• Acquisition of axial cine images at normal and low resolutions.
• Reconstruction of low-resolution cine data using a DL-SR framework (cineDL).
• Quantitative assessment of image quality metrics and qualitative scoring; comparison of volumetric and functional analyses.

Main Results:

• cineDL acquisition was significantly faster (134s vs 252s, P<0.001).
• cineDL demonstrated comparable or superior image quality metrics and qualitative scores, with reduced sensitivity to fetal motion.
• cineDL revealed additional cardiovascular structures in 24% and pathologies in 12% of fetuses.
• Volumetric and functional results were comparable between cineDL and standard resolution images.

Conclusions:

• DL-SR reconstructions shorten fetal CMR acquisition times while achieving diagnostic quality comparable to standard images.
• This technique is less sensitive to fetal motion, potentially leading to improved diagnostic findings.
• DL-SR shows promise for enhancing the clinical utility of fetal CMR in prenatal CHD assessment.