A frame work for classifying dermoscopic images using soft computing approaches

Abstract

newline Skin cancer is a particularly lethal disease because the newlinesymptoms are usually only evident after the cancer has become newlineadvanced. At this stage, any treatment to remove the cancerous tissue newlinehas a high mortality and morbidity rate. Thus, early detection and newlinediagnosis is probably the most critical factor accounting for increasing newlinemelanoma patient survival rates. Computer based analysis and diagnosis newlinemay well provide an avenue for the improvement of diagnostic accuracy newlineand the speed with which it can be achieved. They also provide newlineconsistent quantitative results. newlineThis research work develops an Automated Skin Cancer newlineDiagnosis (ASCD) system using three-dimensional Discrete Wavelet newlineTransform (3DWT). The proposed ASCD system is organized into a newlineseries of modules namely 1) Preprocessing, 2) Feature extraction by newline3DWT, 3) Feature selection and 4) Knowledge discovery. In the 1st newlinemodule, the input dermoscopic images are preprocessed using a wide newlinekernel median filters which removes hairs and noises. In the 2nd module, newline3DWT is employed to find the patterns of skin cancer using texture newlinefeatures available in the obtained wavelet sub-bands. In the 3rd module, newlinefeature selection for multi-class is introduced to select the dominant newlinefeatures separately. This stage helps to increase the accuracy of the