The machine-learning algorithm has not been trained to analyse skin that you would not typically consider suspicious. The lesions we have used to train the algorithm were suggestive of skin cancer (e.g. melanoma, BCC and SCC) to clinicians trained in assessing skin cancer.