Development of Combined Genetic and Imaging Approaches for Differentiation of Cutaneous Malignant Melanoma and Benign Melanocytic Nevi

The morbidity and mortality rate is reduced by premature and exact diagnosis of melanoma, which is the deadliest type of skin cancer. Timely identification of melanoma needs extremely complex and subjective test and laboratory samples. It is not insignificant even for experienced dermatologists to identify, so lot of concentration must be given. Finding the difference between melanoma and mole is also an issue in the accuracy of clinical diagnosis of melanoma. Especially, early diagnosis of cutaneous melanoma is very hard for experienced dermatologists. Even though a lot of advanced imaging techniques and clinical diagnostic algorithms such as dermoscopy and the ABCD rule of dermoscopy respectively are available, clinical diagnosis of melanoma becomes very challenging. The accuracy is an issue of distress (estimated to be about 75--85%) especially with oblique pigmented lesions. Quantitative and objective evaluation of the skin lesion is achieved by the above methods with respect to the subjective clinical assessment. An effective diagnosis can be achieved by reducing the viewer variability’s found in dermatologists' examinations. In order to improve some of existing methods and budding new techniques to ease accurate, fast and reliable diagnosis of cutaneous melanoma. In this paper different types diagnostic system of melanoma namely, preprocessing feature extraction, feature selection and classification is explained. The results of feature selection were optimized from advanced classes of classification techniques; namely, Two weighted k-nearest neighbor (k-NN) classifiers (k = 1, 30), a decision tree (DT), and the Random Forest (RF) algorithm are employed. Support Vector Machine has been very effective in computer-based melanoma diagnosis studies in the literature.