Deep learning models for classification of breast cancer receptor expression using mammograms

A study at Osaka City University (Japan) resulted in development of deep learning-based models classifying breast cancer receptor expressions using mammograms. The models discriminated ER-positive and ER-negative with AUC 0.67 (95% CI 0.58-0.76), PgR-positive and PgR-negative with AUC 0.61 (95%-bti 0,53-0.68), and HER2-enriched and HER2 non-enriched with AUC 0,75 (95%-bti 0,68-0.82).