Fig. 10: Interpretable heatmaps of CAMP. | npj Artificial Intelligence

Fig. 10: Interpretable heatmaps of CAMP.

From: CAMP: continuous and adaptive learning model in pathology

Fig. 10

a–d For each row, the name of the task and the prediction of CAMP are shown. The blue texts are similar to the ground truth, whereas the orange text is different from the label. In the first two columns, a WSI and a generated attention heatmap are shown. The heatmaps represent the contribution of tissue patches to the final prediction. We show three important regions at the bottom of each WSI/heatmap, which are directly related to the diagnosis of the slide, e.g., tumors in breast cancer sub-typing. Additionally, patches with high and low scores are depicted in the last two columns for detailed observation.

Back to article page