A Dermoscopic Inspired System for Localization and Malignancy Classification of Melanocytic Lesions

This study aims at developing a clinically oriented automated diagnostic tool for distinguishing malignant melanocytic lesions from benign nevi in diverse image databases. Due to the presence of artifacts, smooth lesion boundaries, and subtlety features, accuracy such systems gets hampered. Thus, proposed framework improves melanoma detection by combining clinical aspects dermoscopy. Two methods have been adopted achieving aforementioned objective. Firstly, artifact removal localization are performed. In second step, various significant features as shape, color, texture, pigment network detected. Features further reduced checking their individual significance (i.e., hypothesis testing). These feature vectors then classified using SVM classifier. specific domain used this design opposed abstract images. The knowledge an expert enhanced methodology. approach is implemented on multi-source dataset (PH2 + ISBI 2016 2017) 515 annotated images, thereby resulting sensitivity, specificity 83.8%, 88.3%, 86%, respectively. experimental results promising, can be applied detect asymmetry, network, colors, texture lesions.