How can artificial intelligence models advance placental biology?

Area of Science:

• Reproductive Biology
• Computational Pathology
• Medical Imaging Analysis

Background:

• The placenta is crucial for fetal development, and its histologic examination reveals functional abnormalities.
• Digital pathology and machine learning (ML) offer powerful tools for analyzing complex tissue image datasets.

Purpose of the Study:

• To explore the application of ML models in placental histology for identifying structural and functional abnormalities.
• To highlight the utility of ML in supplementing pathologist expertise and generating new research hypotheses in placental pathology.

Main Methods:

• Development and application of cell classification ML models on placental disc and membrane tissues.
• Utilizing small image patch analysis and aggregation methods for whole-slide characterization.
• Focus on feature extraction from digital pathology slides to derive biological and clinical insights.

Main Results:

• ML models successfully perform cell classification, offering a 'bottom-up' approach to tissue characterization.
• Identified pathologies within great obstetric syndromes and placental inflammation by aggregating findings from image patches.
• Extracted features from digital pathology slides provide valuable biological and clinical knowledge.

Conclusions:

• Machine learning models are effective tools for analyzing placental histology and identifying deviations from normal development.
• ML applications in digital pathology enhance the understanding of placental function and disease, supporting research and clinical insights.