Fig. 5.

Representative images from a level 1 case with an incorrect prediction: (a) the original image was labeled “Heel”; (b) the prediction score = 0.9665 indicates a high level of confidence in the incorrect classification of the label “Ankle” for the image; (c) although this image was labeled correctly as “Heel”, the prediction score = 0.1594 indicates a low level of confidence in the prediction. The prediction in (b) is an example of the advantage of “Allowed Labels” for improving performance. “Ankle” is an “Allowed Label”, for “Heel”, which improves performance when calculating “AllowedLabel_Sensitivity”