Decidual Vasculopathy Identification in Whole Slide Images Using Multiresolution Hierarchical Convolutional Neural...

Area of Science:

• Perinatal Pathology
• Computational Pathology
• Artificial Intelligence in Medicine

Background:

• Placental examination is crucial for infant health assessment, identifying conditions like infection and maternal vascular malperfusion.
• Decidual vasculopathy (DV) is a placental lesion associated with adverse pregnancy outcomes, including preeclampsia, impacting maternal and neonatal health.
• Current placental inspection is limited by high delivery volumes and the need for expert pathologists, leading to many placentas being discarded without examination.

Purpose of the Study:

• To develop an automated, hierarchical machine learning approach for detecting and classifying decidual vasculopathy (DV) in digitized placenta slides.
• To integrate image features with patient metadata for predicting the presence of DV.
• To enable broader, standardized placental screening and identify cases requiring expert pathological review.

Main Methods:

• A hierarchical machine learning model was developed for automated detection and classification of DV lesions.
• Image features extracted from digitized placenta slides were combined with patient metadata.
• The model was trained to predict the presence of DV based on combined data.

Main Results:

• The study introduces a novel machine learning approach for automated DV detection in placentas.
• The method combines image analysis with patient data for enhanced prediction accuracy.
• This technology facilitates more comprehensive placental screening.

Conclusions:

• Automated analysis of placenta slides can significantly improve the detection of critical lesions like DV.
• This computer-assisted approach allows for real-time adjustments in infant and maternal care.
• Early identification of at-risk pregnancies can guide interventions, such as aspirin therapy, to prevent preeclampsia.