Skip to main content
2025 BMES Annual Meeting Main banner
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.

Women's Health

Women's Health - Poster Session C & D

Z30 - Assessment of Placental Microstructure via Ultrasonic Texture Analysis for Diagnosing Fetal Growth Restriction

Friday, October 10, 2025
10:00 AM - 11:00 AM PDT
Location: Exhibit Hall F, G & H

Introduction: : Fetal growth restriction (FGR) is a condition in which a fetus does not reach its full growth potential in pregnancy, due to compromised placental function [1]. 5-10% of pregnancies are affected by FGR, with FGR being responsible for 30% of stillbirths and a leading cause of intrapartum asphyxia [1]. Currently, there is no gold standard for FGR diagnosis [1]. Clinical methods of predicting FGR involve ultrasonic measurements of fetal size and umbilical artery Doppler blood flow measurements [1]. Although these approaches can point to signs of placental insufficiency, they do not assess placental microstructure, with microstructural abnormalities of the placental vasculature known to be related to placental complications [2, 3]. Additionally, fetal size and umbilical artery blood flow measurements require specialized training and may not be feasible to collect in low-resource settings. Therefore, there is a need to develop a method to assess placental structure and function that can be widely implemented. In this work, we present the development of an approach to analyze the placenta in routine ultrasonic images using texture analysis and its implementation in a clinical cohort.

Materials and

Methods: : A total of 122 participants were recruited to this IRB-approved study, with placental ultrasounds collected at two gestational ages: 28–32 weeks and 34–38 weeks. There were four outcome groups: severe FGR (baby birthweight less than the third percentile, N = 10), Stillbirth near-miss (babies with a high risk of stillbirth as previously described [2], N = 14), non-FGR (baby birthweight greater than the tenth percentile, N = 64), and other (did not meet criteria of other groups, N = 34). The placenta was manually segmented from the ultrasound images using ITK-Snap to generate masks [4]. The texture of the placental ultrasound image was evaluated using PyRadiomics, an open-source python package for radiomics analysis [5]. To optimize the texture analysis pipeline, several sensitivity studies were performed. This included mask border erosion to minimize the effects from mask imperfections, removal of acoustic shadows from fetal bones over posterior placentas, image pre-processing, and determining the effect of ultrasonic settings on texture results. Then, with the final developed texture analysis pipeline, the placental ultrasound images in the severe FGR and non-FGR cohorts were compared using a linear mixed model, with a P< 0.01 considered significant.

Results, Conclusions, and Discussions:: Texture analysis of placental ultrasound images demonstrated statistically significant differences between severe FGR and non-FGR cases. For images collected between 28–32 weeks, only one texture feature was found to be significantly different between severe FGR and non-FGR. However, at 34–38 weeks gestation, twenty significantly different features were identified (Fig. 1A). How these differences in texture reflect placental microstructure is still being investigated, but these results demonstrate a promising tool that could be clinically implemented for FGR diagnosis. To promote future clinical implementation, various sensitivity studies were conducted to guide pipeline development. A border erosion level of 20% of the original area was selected because changes were found between all tested levels of erosion, with 20% of the original area being the lowest tested that would account for uncertainty at the placental border. For removing the acoustic shadows from the hand-segmented masks, it was found that a static pixel value threshold of 70 on a normalized placental image was adequate to identify the shadowed regions in the mask. Furthermore, equalizing the image itself had the best performance for distinguishing severe FGR from non-FGR placental images. With the pre-processing steps finalized (Fig. 1B), the effects of ultrasonic settings were analyzed. Image scale was significantly associated with texture feature values, with differences between severe FGR and non-FGR images decreasing with increasing image scale (Fig. 1C). Therefore, image scale was included as a covariate in the linear mixed model. The work presented here establishes a novel screening tool for severe FGR, which could be used in a variety of clinical settings.

Acknowledgements and/or References (Optional): : Acknowledgements: 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.

References: [1] Nardozza et al., Arch Gynecol Obstet, 295:1061-1077 (2017). [2] Gordijn SJ et al., Ultrasound Obst Gyn. 48(3):333-339 (2016). [3] Kingdom J et al. J Obst Gynaecol Canada. 45(10):102154 (2023). [4] Yushkevich PA et al. Neuroimage. 31(3):1116-1128 (2006). [5] van Griethuysen JJM et al. Cancer Res. 77(21): e104–e107 (2017).