Fig. 3: HER2 tumor status classification using annotated slides. | Modern Pathology

Fig. 3: HER2 tumor status classification using annotated slides.

From: Deep learning trained on hematoxylin and eosin tumor region of Interest predicts HER2 status and trastuzumab treatment response in HER2+ breast cancer

Fig. 3

AUC/ROC for HER2 status classification at the slide-level (a) and at the tile-level (b) for both transfer learning and the fully trained models. c Tile-level AUC of the three-way classifiers. d Two representative H&E slides from the test set with corresponding heatmaps based on predicted probabilities by the CNN model. Yellow regions indicate the ROIs determined by our pathology team. The middle panel shows predicted heatmaps for unannotated model and the right panel shows the predicted heatmaps from three-way annotated model.

Back to article page