Development of dynamic image analysis methods to measure vascularisation and syncytial nuclear aggregates in human placenta

Abstract

Histological examination of the placenta significantly contributes to diagnosis in adverse birth outcomes. One challenge in image analysis is variation in staining intensity caused by batch variation. We investigated if dynamic threshold image analysis methods may increase accuracy. Placenta samples were stained for endothelial cells and syncytial nuclear aggregates and analysed in Qupath software. Dynamically setting the threshold resulted in data more similar to manual method data. The method is simple and effective at modelling the dynamic interpretation of variation in staining intensity achieved by manual methods. We anticipate dynamic methods could be used to enhance placental diagnosis.

Keywords: Adaptive threshold; Image analysis; Immunohistochemistry; Machine learning.