Detection and classification of skin lesion based neural networks

Abstract

Skin cancers emerge from the skin due to the growth of untypical cells that dispersal to other parts of the body. Through early detection, very high success rates can be achieved in treating most cases of skin cancer, including the most aggressive and deadly forms and types. The objective of this work is brief as follows: First, to collect a real database from Al-Kindi Hospital and Baghdad Medical City. Next, choose smart classification methods for the early detection of skin lesions. Second, propose a new technique for reducing database noise obtained based on convolutional neural networks (CNN). Third, a new algorithm has been proposed for dividing a cutaneous lesion into cutaneous images based on the straight active-contour and morphological processes.