FIGURE 5:

Class activation maps (CAMs) highlight which pixels in the images are important for the model to make its classification probability. The darker areas of the map show where the network focused its decision-making on. The three images are the same slice from one patient in the test set. (Left) Map from the cartilage lesion detection network. (Middle) Map from the bone marrow edema-like lesion detection edema detection network. (Right) Map from the subchondral cyst-like lesion detection network.