Ultrasound Texture Analysis Correlates to Placental Microvascular Structure in Fetal Growth Restriction and Healthy Pregnancies

Introduction: : Fetal Growth Restriction (FGR), affecting 5–10% of pregnancies, is the leading cause of perinatal mortality [1]. Fetal growth depends on the exchange efficiency of the placental villous microvasculature, where oxygen and nutrients are transported from the maternal blood to the fetal capillaries contained within the villous tissue [2]. However, there are currently no tools routinely used to detect changes in structure of the placenta villous tissue in vivo because the resolution of standard-of-care ultrasound (US) is limited. Instead, risk for severe FGR that could lead to stillbirth is assessed typically through estimated fetal weight and Doppler US, yet 40% of all stillbirths are considered preventable [3]. Therefore, there is a need to develop a tool to assess microvascular changes in the placenta, which could be implemented in routine clinical care. Texture analysis, or the quantification of pixel intensities and the relationship between pixels in an image, presents a promising method to detect changes in structure with US imaging [4]. However, it is unknown if texture analysis could capture changes at a length scale of the microvasculature of the villous tissue. Therefore, the objective of this work was to characterize the relationship between texture analysis of placenta US images and microvascular structure assessed through microscopy techniques after birth.

Materials and

Methods: : Participants were recruited under an IRB approved study and US images of the placenta were acquired around 36 weeks (± 2 weeks) of gestation (N = 100) (Figure 1A). Patients with severe FGR were defined as patients whose babies had a birthweight less than the third percentile (N = 11) [5]. Placentas from acquired B-mode US images were manually segmented and the placental image texture was evaluated using PyRadiomics [4] (Figure 1B). To assess microvascular changes in the placenta, placentas from each patient were collected after birth. Biopsies of placenta villous tissue were fixed in 10% neutral buffered formalin. Fixed tissue biopsies were partially optically cleared with Visikol Histo-1 and imaged with Optical Coherence Tomography (OCT) (Figure 1A,C) [2]. Three-dimensional images (2 mm x 2 mm x 1.5 mm) were segmented, 3D reconstructed (Figure 1D) and quantified with a custom MATLAB code to measure surface area per volume (SA/V), villi thickness (t), and volume fraction (VF) (N = 18 images/patient, N = 100 patients). To compare changes in structure of OCT images acquired, the severe FGR patients (N = 11) were gestationally age-matched to an equal number of controls (N = 11). A linear mixed effect model with a one-way ANOVA was used to determine statistical significance. Subsequently, Pearson’s correlation coefficients were calculated for the texture features and the average OCT measurements for each patient (N = 100) and statistical significance of the correlations was assessed.

Results, Conclusions, and Discussions:: Quantification of OCT images revealed alterations in placenta villous tissue between severe FGR and control placentas (Figure 1E). Surface area per volume (SA/V) of placenta villous tissue decreased with severe FGR, indicating a decrease in surface area available for the exchange of oxygen and nutrients. Additionally, the villi thickness increased with severe FGR. This increase in villi thickness is consistent with the literature [6] and also suggests greater diffusion distance from the maternal blood to the fetal capillaries, which would also decrease the gas exchange. When comparing OCT measurements (surface area per volume, villi thickness, and volume fraction) to texture analysis features (e.g., First Order Skewness, GLCM Sum Entropy, GLRLM Gray Level Non-Uniformity), 41 texture features significantly correlated with OCT measurements (p < 0.05). Select features with the greatest Pearson correlation coefficients to OCT measurements are shown (Figure 1F). Future work will explore the capacity of texture analysis to detect known changes in the microstructure of ultrasound phantoms. Additionally, texture analysis features will be used to build predictive models of severe FGR to understand when clinical intervention (e.g., early delivery) is necessary to prevent stillbirths. Overall, this work suggests that implementing an advanced image analysis technique to quantify US images of the placenta presents a promising method to assess changes in underlying placental structure associated with severe pregnancy complications.

Acknowledgements and/or References (Optional):: The authors thank all individuals who participated in this study. Work on this project is supported by Wellcome Leap as part of the In Utero Program.
[1] Nardozza, L.M.+ Arch. Gynecol. Obstet, 2017; [2] Scott A.K. + Placenta, 2024; [3] www.who.int; [4] Griethuysen J.J.M. +, Cancer Res. 2017; [5] Papageorghiou, A.T. + AJOG, 2018; [6] Baergen, R.N + Benirschke’s Pathology of the Human Placenta, Springer Nature, 2022.