Fetal brain biometry and sulcal depth in uncomplicated pregnancies: an interrater correlation coefficient study in low- versus super-resolution MRI

Objectives

Cortical fetal brain MRI provides valuable insights into neurodevelopment. Low-resolution (LR) images pose reorientation problems in orthogonal planes. Hence, the use of super-resolution (SR) reconstructions from LR sequences offers better spatial resolution and allows unrestricted 3D navigation. We aim to assess the interrater reproducibility of SR and its ability to improve measurement reliability.

Methods

We obtained 36 fetal brain MRI sequences from uncomplicated pregnancies between 26-37 weeks’ gestation. Brain reconstructions were generated from a minimum of four stacks of T2w acquisitions in at least one for each orthogonal plane. 3D SR volume reconstruction was done using NiftyMIC. Measurements were performed following ISUOG criteria for: Biparietal diameter (BPD), occipito-frontal diameter (OF), cerebellum, corpus callosum (CC) and vermis. Sulcal depths were measured according to Hahner et al (2019). Both SR and LR datasets were measured by two raters using Horos and ITK-snap software, respectively. Interrater correlation coefficients (ICC) were employed for quantitative analysis of variations. Scores < 0.4 were classified as poor, 0.4-0.59 as fair, 0.6-0.74 as good, and >0.75 as excellent.

Results

ICC outcomes within the SR group revealed excellent correlation for the CC (0.79), vermis (0.92), cerebellum (0.98), BPD (0.91), OF (0.87), insula (0.9), right parieto-occipital (0.77), and bilateral calcarine sulci (0.87). Good correlations were observed for the left parieto-occipital sulcus (0.62) and cingulate sulci (0.7). Conversely, LR sequences had excellent interrater correlations for cerebellum (0.98), BPD (0.98), and OF (0.85), and good correlations for the CC (0.62). However, correlations for sulcus measurements ranged from low to poor (0.08-0.48).

Conclusions

SR reconstructions provide a promising approach for quantifying fetal brain MRI, enhancing reproducibility and ensuring precise alignment of planes, especially for sulcation measurements.