Fabio De Martino

CNN patch classifiers trained only on normal breast H&E slides label epithelium, stroma and adipocytes with AUCs of 0.98–1.00 on three external cohorts and feed a lobule-detection pipeline.

The paper introduces a deep learning tool to classify xenograft cells in histological images by analyzing contextual features, achieving up to 96% accuracy in distinguishing human and murine cells.