Introduction
The antenatal work-up of fetal thoracic anomalies increasingly includes fetal MRI in recent years. Fetal MRI aids in cases of diagnostic uncertainty by detailed delineation of soft tissues within the fetal chest, differentiating lung, bowel and liver more easily than ultrasound1, helping in the assessment of congenital diaphragmatic hernia (CDH) and congenital lung lesions (CLL, including congenital pulmonary airway malformation [CPAM] and bronchopulmonary sequestration [BPS]). The currently standardised prognostic markers for CDH and CPAM are the observed/expected Lung Head Ratio (O/E LHR,2) and the CPAM volume ratio (CVR,3), both measured by ultrasound. Both of these metrics utilise single plane measurements of aspects of the fetal chest to extrapolate volumes. Although these metrics are valid surrogates for total lung volume (TLV, 4), generation of such values may have inherent intra- and inter-operator variability5. Discrepancies between 2D ultrasound and MRI-derived TLV measurements have been shown to be principally related to the failure to account for the contribution of the ipsilateral lung6. Furthermore, ultrasound-based imaging struggles with maternal habitus and fetal positioning, as well as extremes of liquor volume. 3D ultrasound lung volume measurements have been shown to be more difficult than by MRI, mainly because the most hypoplastic lung cannot be properly visualized7. High volume fetal medicine centres have suggested that MRI-calculated lung volumes may be a more accurate predictor of survival but that has not been formally proven8, however there is no standardised methodology for the use of MRI in these cases which limits the compilation of large datasets necessary for further study9.
Modern clinical fetal MRI protocols (primarily based on single shot turbo spin echo (ssTSE) sequence) allow fast acquisition of individual 2D slices that “freeze” fetal position in time. These slices have sufficiently high image and contrast resolution for diagnostic purposes in cases where fetal motion may previously have limited the information available. However, the misalignment between individual slices leads to corruption of volumetric information and loss of structural continuity within a slice stack. Therefore, output MRI stacks are termed “motion corrupted”. In general, to achieve a sufficient coverage of fetal visceral organs, clinical examinations require 3-6 MRI stacks acquired under different orientations with respect to the fetal body. The degree of motion corruption (misalignment of slices) can vary between the stacks and normally depends on the gestational age, amount of amniotic fluid, as well as fetal lie and mobility. It directly influences the accuracy of the fetal lung volume assessment, which is based on 2D slice-wise segmentation of motion-corrupted stacks followed by 3D interpolation. Therefore, calculated TLV values might vary for different stacks (from the same acquisition / the same subject) depending on the amount of motion corruption and acquisition plane. This is considered one of the limiting factors for MRI-based TLV assessment.
The recently proposed deformable slice-to-volume registration (DSVR) method 10 is used for reconstruction of high-resolution (e.g., 0.8x0.8x0.8mm) isotropic 3D images of fetal body from multiple low-resolution (e.g., 1.25x1.25x1.25mm) motion-corrupted stacks. In DSVR method, one of the low-resolution stacks is selected as an initial target space and it is then registered to each of the slices using nonlinear free form deformation registration. This is followed by super-resolution reconstruction (SR) of the 3D image from the registered slices. The full pipeline includes 3 interleaved SR and SVR steps. The resulting reconstructed images provide detailed 3D volumetric information and can be reoriented in any plane. This facilitates accurate 3D segmentations of fetal lungs and other organs allowing volumetric analysis.
In this work, we sought to explore how DSVR-derived lung volumes would compare to those calculated by conventional manual 2D segmentation in cases of fetal thoracic anomalies, with a comparison to normal control cases.