Fig. 4: HEMnet performance on unseen H&E slides. | npj Precision Oncology

Fig. 4: HEMnet performance on unseen H&E slides.

From: A deep learning model for molecular label transfer that enables cancer cell identification from histopathology images

Fig. 4: HEMnet performance on unseen H&E slides.

a Prediction of p53 stain pattern on 17 unseen H&E slides. b Prediction of cancer regions on 13 unseen H&E slides compared to pathologist annotations. c Prediction performance of p53 stain pattern is positively correlated with the ability of p53 to mark cancer regions on the tissue, as annotated by pathologist. d HEMnet accurately predicts cancer regions annotated by pathologist (bottom) and p53 stain (top) when p53 stain agrees with pathologist annotation. e HEMnet predicts cancer regions even when p53 stain pattern (left) disagrees with pathologist ground truth annotations (right).

Back to article page