Association of Deep Learning-Derived Histologic Features of Placental Chorionic Villi with Maternal and Infant...

Quantification of placental histopathological structures is challenging due to a limited number of perinatal pathologists, constrained resources, and subjective assessments prone to variability. Objective standardization of placental structure is crucial for easing the burden on pathologists, gaining deeper insights into placental growth and adaptation, and ultimately improving maternal and fetal health outcomes.

Leveraging advancements in deep-learning segmentation, we developed an automated approach to detect over 9 million placenta chorionic villi from 1,531 term placental whole slide images from the New Hampshire Birth Cohort Study. Using unsupervised clustering, we successfully identified biologically relevant villi subtypes that align with previously reported classifications - terminal, mature intermediate, and immature intermediate - demonstrating consistent size distributions and comparable abundance. We additionally defined tertile-based combinations of villi area and circularity to characterize villous geometry. This study applies these cutting-edge AI methods to quantify villi features and examine their association with maternal and infant characteristics, including gestational age at delivery, maternal age, and infant sex.

Increasing gestational age at delivery was statistically significantly associated (p=0.003) with an increase in the proportion of mature intermediate villi and a decrease in the proportion of the smallest, most circular villi (p < 0.001). Maternal age and infant sex were not statistically significantly associated with measures of villous geometry.

This work presents a workflow that objectively standardizes chorionic villi subtypes and geometry to enhance understanding of placental structure and function, while providing insights into the efficiency, growth, and the architecture of term placentas which can be used to inform future clinical care.