Results of class activation mapping on a patient with pure DCIS. Input image showing heterogeneous calcifications (a). Guided backpropagation (b), highlighting every pixel used in making the decision for each class, whether that pixel was a negative or positive predictor. In GRAD-CAM, the areas that are highlighted are the regions that were a positive factor in predicting the specific class (c). In Guided Grad-CAM (d), previous methods are combined to highlight pixels the network pays attention to in order to provide positive inputs, which include the region of calcifications as well as intervening breast parenchyma.