Production of the Grounds for Melanoma Classification Using Adaptive Fuzzy Inference Neural Network

Several researchers investigated automated diagnosis for malignant melanomas as known as the worst skin cancer. Those systems, however, only provide final discrimination results but not related information such as their substantial reasons and therefore reliability of the system still remain an open issue. In this paper, we developed a new melanoma screening system based on an adaptive fuzzy inference neural network (AFINN). Our new system provides not only final discrimination result but also its grounds in easy-to-read fuzzy if-then format Our system developed 88 fuzzy rules in consequence of the learning of 1148 dermoscopy images and in the classification, it provides both of the final result and its constituent rules. Based on only developed rules, our system achieved a sensitivity of 81.5% and a specificity of 73.9%. Since it is almost equivalent to expert dermatologists', we consider the developed rules are reasonable and this supplemental information improves overall system reliability.