Fusion of Color Bands Using Genetic Algorithm to Segment Melanoma

Abstract

Melanoma is often associated with changes in the color, size or shape of a mole. Several tools, including smartphone apps, have been developed for the detection of melanoma through medical images. To interpret the information in these images efficiently, it is necessary to isolate key regions of interest. In this research, we analyze the impact of evolutionary computing on the segmentation of melanoma through the fusion of color bands from the images. The tests performed on the PH2 image base showed a 20% improvement in the average Dice compared to using the standard intensity, a promising result toward obtaining more efficient and accurate melanoma screening.