Explainable AI consists in developing models allowing interaction between decision systems and humans by making the decisions understandable. We propose a case study for skin lesion diagnosis showing how it is possible to provide explanations of the decisions of deep neural network trained to label skin lesions.