Fetal Brain Biometric Measurements on 3D Super-Resolution Reconstructed T2-Weighted MRI: An Intra- and Inter-observer Agreement Study

Marie Khawam, Priscille de Dumast, Pierre Deman, Hamza Kebiri, Thomas Yu, Sébastien Tourbier, Hélène Lajous, Patric Hagmann, Philippe Maeder, Jean‐Philippe Thiran, Reto Meuli, Vincent Dunet, Meritxell Bach Cuadra, Mériam KoobFrontiers in Pediatrics 2021 · DOI

Abstract

This study evaluates the use of 3D super-resolution (SR) reconstructed volumes for fetal brain biometry compared to traditional 2D T2-weighted sequences (T2WS). Analyzing 25 normal fetal MRI scans (18–34 weeks), researchers assessed the level of confidence and concordance between an experienced pediatric radiologist and a junior radiologist. Results showed that SR allows for similar biometric measurements while significantly improving diagnostic confidence, particularly for junior observers, by providing a motion-free, navigable 3D volume.

Introduction: The Challenge of Fetal Motion

Fetal brain biometry is a cornerstone of prenatal diagnosis, yet it remains technically challenging due to constant fetal motion. While ultrasound is the primary screening tool, MRI provides superior soft-tissue contrast for detailed follow-up. However, standard clinical MRI protocols rely on 2D slices that are often misaligned with anatomical axes because the fetus moves during acquisition. Radiologists must mentally synthesize these disjointed images, a task that requires extensive expertise and often leads to uncertainty, especially when measuring complex structures like the corpus callosum.

Methodology and Super-Resolution Reconstruction

The study utilized 25 normal fetal MRI scans. The technical workflow involved acquiring orthogonal 2D T2-weighted HASTE sequences and processing them through the MIALSRtoolkit, an open-source pipeline. This process involves fetal brain extraction, bias field correction, and slice-to-volume registration to correct for motion. The final output is an isotropic 3D volume (approx. 1.1 mm³ resolution) that provides a motion-free representation of the fetal brain.

The transition from motion-corrupted 2D clinical acquisitions to a motion-free 3D Super-resolution (SR) reconstruction.

Two observers—an expert (15 years experience) and a junior radiologist—performed 11 biometric measurements (e.g., biparietal diameter, cerebellar diameter) on both the original 2D slices and the 3D SR volumes to compare accuracy and observer confidence.

Dataset and Subject Distribution

The cohort consisted of 25 subjects ranging from 18 to 34 weeks of gestational age. The reconstruction process utilized multiple series per subject to ensure high-quality output, with a mean of 6.04 series used per reconstruction.

Subject distribution by gestational age and the number of MRI series utilized for the SR reconstruction.

Key Results: Closing the Confidence Gap

The study found excellent agreement between measurements taken from 2D slices and 3D SR volumes. Lin’s concordance correlation coefficient (CCC) and Intraclass Correlation Coefficient (ICC) averaged 0.997 for both raters across most metrics.

The most significant impact of SR was on observer confidence. The junior radiologist reported "low confidence" in 52% of 2D cases, primarily due to motion artifacts and poor plane alignment. With 3D SR volumes, this figure dropped to just 8%. SR allows for Multiplanar Reformations (MPR), enabling clinicians to freely rotate the volume to find the precise anatomical planes required for measurement, effectively "undoing" the effects of fetal motion and making the process less dependent on specialized training.

Discussion and Future Directions

While 3D SR reconstruction improves confidence and standardizes biometry, some limitations remain. The current processing time is approximately one hour, which limits its use to post-exam reporting rather than real-time assessment. Additionally, while the geometry is highly accurate, the "synthetic" nature of SR voxels means their intensities may not yet be suitable for advanced tissue signal analysis. Future work will focus on automating the pipeline for clinical integration and validating the technique on pathological cases such as ventriculomegaly.

Conclusion

3D Super-Resolution reconstruction is a robust tool that brings fetal MRI biometry closer to the ease of use found in 3D ultrasound. By providing a motion-free, navigable volume, it reduces the expertise gap and increases the reliability of prenatal neurological assessments.

Summary updated

Abstract

We present the comparison of two-dimensional (2D) fetal brain biometry on magnetic resonance (MR) images using orthogonal 2D T2-weighted sequences (T2WSs) vs. one 3D super-resolution (SR) reconstructed volume and evaluation of the level of confidence and concordance between an experienced pediatric radiologist (obs1) and a junior radiologist (obs2). Twenty-five normal fetal brain MRI scans (18-34 weeks of gestation) including orthogonal 3-mm-thick T2WSs were analyzed retrospectively. One 3D SR volume was reconstructed per subject based on multiple series of T2WSs. The two observers performed 11 2D biometric measurements (specifying their level of confidence) on T2WS and SR volumes. Measurements were compared using the paired Wilcoxon rank sum test between observers for each dataset (T2WS and SR) and between T2WS and SR for each observer. Bland-Altman plots were used to assess the agreement between each pair of measurements. Measurements were made with low confidence in three subjects by obs1 and in 11 subjects by obs2 (mostly concerning the length of the corpus callosum on T2WS). Inter-rater intra-dataset comparisons showed no significant difference (<i>p</i> > 0.05), except for brain axial biparietal diameter (BIP) on T2WS and for brain and skull coronal BIP and coronal transverse cerebellar diameter (DTC) on SR. None of them remained significant after correction for multiple comparisons. Inter-dataset intra-rater comparisons showed statistical differences in brain axial and coronal BIP for both observers, skull coronal BIP for obs1, and axial and coronal DTC for obs2. After correction for multiple comparisons, only axial brain BIP remained significantly different, but differences were small (2.95 ± 1.73 mm). SR allows similar fetal brain biometry as compared to using the conventional T2WS while improving the level of confidence in the measurements and using a single reconstructed volume.
BibTeX
@inproceedings{khawam2021fetal,
  author = {Marie Khawam and Priscille de Dumast and Pierre Deman and Hamza Kebiri and Thomas Yu and Sébastien Tourbier and Hélène Lajous and Patric Hagmann and Philippe Maeder and Jean‐Philippe Thiran and Reto Meuli and Vincent Dunet and Meritxell Bach Cuadra and Mériam Koob},
  title = {{Fetal Brain Biometric Measurements on 3D Super-Resolution Reconstructed T2-Weighted MRI: An Intra- and Inter-observer Agreement Study}},
  booktitle = {Frontiers in Pediatrics},
  year = {2021},
  doi = {10.3389/fped.2021.639746},
}
Figure