Advanced skin lesion discrimination pipeline for early melanoma cancer diagnosis towards PoC devices

Abstract

The proposed work describes an effective pipeline for skin lesion (nevus) analysis with related oncological outcomes. The increasing statistics of skin cancer have recently contributed to the development of new methods for early detection and discrimination of malignant skin lesions in order to drastically reduce the number of biopsies often very invasive for the patients. The main aggressive skin cancer histology is the so-called “melanoma” with related differentiation. Several methods have been proposed in the literature for early melanoma detection but often they lack in sensibility/specificity for a real clinical use. The proposed pipeline is based on an effective approach employing analytic innovative hand-crafted image features combined with a machine learning system. This allows both early detection and discrimination of the skin lesions with good trade-off between sensibility/specificity ratio.