Characterization of melanomas using a variety of features

In recent years, skin cancer has been observed to be one of the most hazardous form of the cancers diagnosed in humans. Although Melanoma is the most unpredictable type of cancer among various type of skin cancers, early detection of Melanoma cancer can be helpful to cure it. There are severeal criteria that doctors check during histological diagnosis of a risky skin area, namely color distribution, pigment network, symmety/asymmetry of boundary, regression area, streaks and dot/globules. These physical measures that have long been used to check skin cancer, have also caught the attention of researchers that would like to mathematically model the features using image processing and automate recognition using machine learning. The aim of this work is to propose features for regression area and symmety/asymmetry of boundary measures. Elliptic Fourier Descriptors and Radial Distance Function are utilized to model Symmety/asymmetry measure and results are compared to those computed using Jaccard similarity coefficient.