Potential detection application of nodular melanoma on melanocytic nevi image based on android

Abstract

Nodular melanoma is a deadly rare type of skin cancer. Nodular Melanoma has characteristics asymmetrical shape, border irregularity, nonhomogeneous or has several color variations and the diameter is more than 6 millimeters. Nodular melanoma has a physical form similar to melanocytic nevi, therefore nodular melanoma can be detected from melanocytic nevi spread throughout the body. This research aims to detect nodular melanoma through melanocytic nevi by utilizing the android system in order to ease the user by using camera smartphone in detecting cancer. This application uses image processing and feature extraction of the ABCD method to process images with decision tree c4.5 classification method to detect potential of nodular melanoma diagnosis from melanocytic nevi image. The ABCD method is a medical method used to detect the possibility of skin cancer using 4 parameters including asymmetrical shape, border irregularity, color and diameter. Decision tree c4.5 is classification method that using entropy and gain to make rules of decision tree. The image data test is obtained from the results of the android-based smartphone camera shooting and from medical record. Output of this application is a diagnosis condition of melanocytic nevi is healthy or nodular melanoma potentially. The accuracy of this application is 97.5%.