Multiple Instance Learning approaches for Melanoma and Dysplastic Nevi images classification

Malignant melanoma is responsible for the highest number of deaths related to skin lesions. The similarities of melanoma with other skin lesions such as dysplastic nevi, however, constitute a pitfall for early diagnosis. The research community is committed to proposing software solutions that favor the computerized analysis of lesions for melanoma detection. Existing methods have typically focused on the dichotomous distinction of melanoma from benign lesions. Currently, there is debate about Dysplastic Nevi Syndrome (DNS), or rather about the number of moles present on the human body as potential melanoma risk factors. Distinguishing dysplastic nevi from common ones is a challenging yet mostly unexplored classification problem. The classification phase is particularly delicate: over time, a series of automatic learning algorithms have been proposed to better face this issue. In this paper, we refer to the emergencing role of Multiple Instance Learning approaches for discriminating melanoma from dysplastic nevi and to outline the even more complex challenge related to the classification of dysplastic nevi from common ones.