Lung volume segmentation in fetal MRI: super-resolution reconstructions improve inter-rater reliability

Area of Science:

• Medical Imaging
• Radiology
• Fetal Medicine

Background:

• Fetal MRI is crucial for diagnosing fetal lung pathologies.
• Super-resolution (SR) algorithms offer potential for enhanced fetal MRI analysis.
• Improving inter-rater reliability in lung volume measurements is clinically significant.

Purpose of the Study:

• To evaluate if SR reconstructions improve agreement in fetal lung volume measurements between different raters.
• To assess the impact of SR on inter-rater reliability in fetal lung volume assessment.

Main Methods:

• Retrospective analysis of 98 fetal MRI scans.
• Manual segmentation of fetal lungs from original and SR reconstructed images.
• Calculation of fetal lung volumes and coefficient of variation (CV).
• Bland-Altman plots and intraclass correlation coefficients (ICCs) were used to assess agreement.

Main Results:

• SR reconstructions showed significantly lower CV for lung volumes (p < 0.001).
• Higher ICCs were observed for SR reconstructions compared to original images.
• Bland-Altman plots indicated better agreement in lung volumes from SR reconstructions.

Conclusions:

• SR reconstructions of fetal lungs in MRI enhance inter-rater reliability for lung volume assessment.
• SR MRI reconstructions are a valuable tool for consistent fetal lung volume measurements.
• Improved reliability aids in fetal development assessment and outcome prediction.