Comparison between various supervised ANN algorithm using RGB indices for plaque lesion classification

Color and color difference are important information for a lesion in dermatological diagnosis. This paper presents various supervised ANN models for plaque classification using RGB indices. Images are taken from selected skin at the dermatological clinic which the images are captured using digital camera with controlled environment. The analysis of dermatological digital images is performed by measurements of the pixels that represent a segmented cropped area of the skin lesion. The segmentation of the skin image could be accomplished automatically by supervised segmentation algorithms. The proposed models are designed by implementing Levenberg Marquardt feedforward, Radial Basis Function algorithm and backpropagation. Each optimized model are evaluated and validated through analysis of the performance indicators regularly applied in classification models. At later stage, the best performance from these models is compared with other researchers' work in this area, but using different color space such as YCbCr and HSV.