Assessment of brain development and severity of pulmonary hypoplasia
Biometric measurements were made on standard T2-weighted images, and included brain and skull biparietal diameter (BPD) and fronto-occipital diameter (FOD), atrial width and transverse cerebellar diameter (TCD), measured following Garel et al. 13 (figure 1) Head circumference (HC) and extra-axial space percentiles were calculated according to Kyiakopoulou et al . 12 Fetal cortical development was scored using the grading system described by Pistorius et al. 15 and measured following EgaƱa-Ugrinovic et al.16, which we have used and reported previously.17 The following brain regions were scored subjectively (by D.E. supervised by M.A.): frontal, parietal, temporal, mesial, insular and occipital cortex. 15 Selected primary sulci and gyri were graded and/or measured, including the parieto-occipital fissure, the central, calcarine, superior temporal, cingulate sulcus and, for the opercularization, the Sylvian fissure.15, 16 In addition to the Sylvian fissure depth, which reflects the distance between the inner part of the skull and the insular cortex, the insular depth was also measured, i.e. the distance from the midline to the insular cortex.17 The sum of all the graded fissures provides a total grading score for the whole brain as well as for each hemisphere. 18
3D super-resolution reconstruction (SRR) volumes were created from the standard T2-weighted 2D stacks displaying the fetal brain, using NiftyMIC, a publicly available and state-of-the-art SRR algorithm.19 The SRR volumes were automatically segmented for white matter, ventricular system (lateral ventricles, third ventricle, fourth ventricle and aqueduct) and the cavum septi pellucidi and cavum vergae, extra-axial space and cerebellum with manual correction when necessary (D.E. supervised by M.A). A deep learning algorithm for the automatic segmentation of white matter, ventricular system, and cerebellum was used for the first volumes that were processed.20 As the number of volumes segmented for the extra-axial space increases, we trained a new deep learning-based segmentation algorithm21 based on a partially supervised learning method that segments automatically white matter, ventricular system, cerebellum, and extra-axial space. These segmentations were used for volumetric analysis when the quality of the SRR volume allowed further analysis (determined by D.E. supervised by M.A. and L.F.). (Figure 2)
In all cases were the fetus underwent FETO the date of it was noted and the time interval between operation and second MRI was documented. The severity of the pulmonary hypoplasia in fetuses with CDH was assessed on MR images by measuring the right, left and total fetal lung volume (TFLV), the fetal body volume (FBV), liver position, intra-thoracic liver volume and thoracic volume, again measured manually (D.E. supervised by M.A.). From those, we calculated the O/E TFLV ratio22 and the liver-to-thoracic volume ratio (LiTR).23 The latter ratios provide biometric measurements in the index case, that are independent of gestational age and /or fetal weight.