Abstract 5470: Low-magnification deep learning model for rapid HER2 status prediction from H&E whole-slide images.

Key Points

• The aim is to develop a low-magnification deep learning model to predict HER2 status from H&E whole-slide images.
• Developed a deep learning model for low-magnification images
• Utilized the TCGA-BRCA dataset for model training and validation
• Implemented a five-fold cross-validation strategy
• Predicted HER2 status as positive or negative from slide-level data
• Achieved an AUC of 0.728 and F1-score of 0.653
• Outperformed the existing model UNI2 with an AUC of 0.715
• Processed 8.8 whole-slide images per minute, 30 times faster than UNI2
• Required significantly less computational and storage resources