Data
Each fetus was scanned using T2-weighted ssTSE sequence, producing between 6 and 10 stacks in different planes, with varying degrees of motion corruption. In all cases, images were acquired on an Ingenia 1.5 Tesla (Philips) system using T2w ssTSE: TR=15000ms, TE=80ms, voxel size 1.25x1.25mm, slice thickness 2.5mm and spacing 1.25mm. For the initial 16 study cases, 2D slice-wise segmentation of the fetal lungs was performed in ITK-SNAP (http://www.itksnap.org) in the axial stack with respect to the fetal trunk and a total fetal lung volume was calculated using the conventional interpolation approach. These datasets also underwent DSVR reconstruction to 0.85x0.85x0.85mm resolution using SVRTK toolbox (https://github.com/SVRTK/SVRTK) and lung volumes were calculated after automated 3D segmentation followed by manual refinement in 3D Slicer (https://www.slicer.org). An example of one of the he original 2D axial stacks affected by minor to moderate motion and the corresponding 3D DSVR reconstruction are displayed side by side for descriptive purposes in Figure 1 and Supplementary Video 1. The processing of the additional 100 cases included reconstruction using the standard DSVR pipeline 10 followed by semi-automated lung segmentation.
Lung volumes from the 16 cases derived from 3D DSVR and the conventional 2D-based approach underwent pairwise Bland-Altmann comparison to assess for bias in measurement. In cases of CLL, lesion volume was calculated and is provided here in cm3 and as a proportion of TLV. A normal curve (growth chart) for fetal lung volumes was generated for 21-32 weeks GA range from 100 DSVR reconstructions , and this curve was compared descriptively to the previously published MRI-generated fetal lung volume normograms 11,12.