Shape and color feature based melanoma diagnosis using dermoscopic images

In this paper, an essential system to identify the melanoma in skin at an early stage is proposed. Skin Cancer (SC) is one of the deadliest disease and its morality rates is very high. A SC classification model is designed based on the novel Color, Shape feature extraction and Classifier to detect the Melanoma which is known as CSC-Mel identification model. In preprocessing, feature and gradient adaptive of contour model is employed to segment the skin lesion. Along with ABCD rule, a novel shape and colour features are extracted as features and K-Nearest Neighbor (KNN) classification is employed for the classification. The CSC-Mel Identification Model is tested on PH2 dermoscopic image dataset with 3-Fold Cross Validation (FCV) for testing and development process. Results shows that the CSC-Mel identification model identifies the skin cancer with an accuracy of 90.5%.