Fig. 5
From: Congo red fluorescence enhances digital pathology workflow in cardiac amyloidosis
Examples of tile classification with their respective visualizations. From left to right, each set includes: the original image, the classification model’s heatmap highlighting areas most considered (warm colors) and least considered (cool colors) for the decision, the ground truth mask from the segmentation model, and the predicted mask from the segmentation model. In negative cases, the heatmap interpretation consistently highlighted large, nonspecific regions, suggesting a broader area of focus in respect to positive cases where amyloid was pointed out by the heatmap.
