Synopsis

Keywords: Fetal, Neuro, gyrification

Current clinical assessment of fetal brain gyrification is performed manually and subjectively. Previous studies that suggested methods for quantitative assessment of fetal brain gyrification were based on 3D reconstruction. In this study we propose an automatic method for quantifying the gyrification from routinely acquired 2D clinical data. We calculate five different gyrification indices for both the right and left hemispheres. We present developmental curves of 134 control fetuses in a wide range of GA (18-37 weeks) and demonstrate differences between control fetuses and fetuses with malformation of cortical development: five with lissencephaly and 13 with polymicrogyria.

Introduction

Normal development of the fetal cerebral cortex has a unique spatio-temporal pattern along gestational age (GA)^1–5, with gyrification starting from the second trimester and continuing during post-natal life5. Malformations of cortical development (MCD) were shown to be associated with worse neurodevelopmental outcomes^6–8. Two common MCDs are lissencephaly (LIS) and polymicrogyria (PMG), both resulting from abnormal neuronal migration and organization⁹.Fetal MRI is increasingly being used for cortical development assessment to support ultrasound or to reveal new findings. To date, using both ultrasound and MRI, the cortical surface is assessed qualitatively by visual inspection. However, some pathologies may be missed, especially when the cortical abnormality is only mild and symmetrical¹⁰. A few studies suggested quantitative measures of fetal cortical folding based on 3D reconstructed cortical surface^11–15. However, these studies were limited by using prospective specific data for 3D reconstruction, developed on small cohorts of mainly normal fetuses, and proposed developmental curves along narrow ranges of GA^11–15. Our aims were: 1) develop a method for quantitative assessment of the gyrification of fetal brain based on routinely acquired clinical 2D MRI data; 2) quantify normal gyrification and symmetry on a large cohort of control fetuses; 3) measure gyrification alternations in fetuses with LIS and PMG.

Methods

Participants and MRI protocol: Coronal T2 weighted MRI of 152 fetuses was collected retrospectively: 134 control fetuses (having normal imaging findings and without any congenital abnormalities), five with LIS and 13 with PMG. MRI was acquired on two 3T scanners from two vendors (GE/Siemens), with various sequences FRFSE, TRUFI, HASTE and FIESTA, and acquisition parameters.
Image processing pipeline: (Figure 1) it consists of 4 stages: A) Brain detection: computation of the brain region of interest¹⁶; B) Brain component segmentation with an existing method¹⁷; manual corrections were performed when needed to ensure accurate contour segmentation; C) Computation of the gyrification index (GI) based on the bounded area between the cerebral cortex and the convex hull of the cortex (GIA, Eq. I); and D) Computation of the GI based on the contour (GIC, Eq. II). $$Eq. I: GIA = \frac{S_{\text{bounded area}}}{{S_{\text{hemisphere area}}}}$$ $$Eq. II: GIC = \frac{l_{\text{cerebral contour}}}{{l_{\text{convex hull contour}}}}$$
A total of five quantitative parameters were extracted: Area-based: (1) Mean GIA=the mean value of the GIA calculated from all volume slices; (2) max GIA=the maximum GIA value from all volume slices; (3) Bounded volume=the sum of areas in all slices between the cerebral brain and its convex hull, normalized to the hemisphere’s volume. Contour-based: (4) Mean GIC; (5) Max GIC. Each parameter was calculated separately for the right and left hemispheres. In addition, brain symmetry indices were calculated for each of the five parameters (Eq. III): $$Eq. III: SI = \frac{Right-Left}{{Right+Left}}$$
Statistical analysis: Changes with GA in the control group were assessed by fitting a second-degree polynomial using GAMLSS^18,19 with BCPE distribution. Comparison between control fetuses and LIS and PMG were performed using Wilcoxon’s rank sum tests, with control for GA. All p-values were adjusted for multiple comparisons by Benjamini-Hochberg²⁰. The 95% nonparametric confidence intervals²¹ were adjusted to a false coverage rate of 0.05¹⁸.

Results

Control: Significant changes in all GI parameters with GA were detected. Figure 2 shows developmental curves²². While most parameters demonstrated increased values with GA, a few showed parabolic shape with a minimum value around GA = 25, which may be explained by changes in the shape of the Sylvian Fissure. No brain asymmetry was detected in the control group for any of the GI parameters (Figure 3).
MCD: Table 1 lists the adjusted p-values and confidence intervals of MCDs compared with control fetuses. Seven parameters were significantly lower in LIS fetuses and four parameters were significantly lower in PMG fetuses. The difference between control and LIS fetuses increases as GA advances.In addition, three fetuses referred to MRI due to suspected LIS in ultrasound were determined with normal cortical development based on MRI reading, supported by normal GI values. No brain asymmetry was detected in LIS and PMG (Figure 3) based on group comparisons.

Discussion

This study presents quantitative parameters of fetal gyrification along GA, applicable for clinical routine 2D MRI data. Changes in all gyrification parameters were detected along GA in the control group corroborating previous studies^11,23,24. No brain asymmetry was detected in control fetuses in all parameters, in line with previous studies which used 3D GI^{11–14,25–30}. 2D and 3D GI may hinder small differences known in specific sulci^1,23,31. Our results show a significant difference between control and LIS fetuses and supported radiological interpretation in three normal fetuses suspected of LIS. Only one previous study¹⁵ quantified the gyrification of three PMG fetuses and reported no differences. Our results in PMG fetuses did show some differences, yet studies of additional gyrification parameters are needed.

Conclusion

This study presents for the first time an automated method for GI quantification based on 2D clinical routine MRI. Quantifying gyrification may allow more precise assessment, revealing subtle changes, therefore assess severity in fetuses with MCD. Multi-parametric analysis of brain gyrification may provide complementary information for fetuses with different pathologies.