Fig. 4
From: Congo red fluorescence enhances digital pathology workflow in cardiac amyloidosis
Applications of AI: A mask annotating amyloid deposits was created and overlayed onto the brightfield CR image, enabling extraction of original tiles and their respective masks. In the first experiment (left), positive and negative original tiles were used to train a classifier (based on ResNet18) to recognize the presence or absence of amyloid deposits. Explainability algorithms evaluated areas most influential in the model’s decision-making process. On the right, the segmentation experiment utilized original tiles and corresponding binary masks to train a UNet-based model.
