Hybrid genetic algorithm — Artificial neural network classifier for skin cancer detection

Abstract

Skin cancer is a deadly form of cancer, affecting skin. It is one of the very deadly form of cancer. It occurs on the melanocytic cells of skin. So skin cancer is also known as Melanoma. Due to the malignancy feature it is also known as malignant melanoma. Melanoma causes the abnormal growth of melanocytic cells which produces the sun protective pigment melanin. Due to that reason melanoma appears as black or brown in color. Main factors that cause melanoma are over exposure to sunlight and genetic factors. Ultraviolet (UV) radiation from the sun is important for maintaining vitamin D levels in the body, but too much UV can cause sunburn, premature ageing, skin and damage and ultimately skin cancer. Early detection can cure skin cancer completely. The probability of cure decreases if it is not treated at early stages. Diagnosing methods like Biopsy, SIAscopy are painful and time consuming. An economical, less time consuming, painless Computer Aided Skin Cancer detection is proposed here. This methodology uses Digital Image processing and Artificial intelligence for skin cancer detection. This methodology involves no direct contact with skin. Only the dermoscopic image is used here. The image after certain image processing techniques is subjected to segmentation. After segmentation, the unique features are extracted from the image using feature extraction techniques. The feature extraction technique used here is GLCM(Gray Level Co-occurrence Matrix) and RGB color feature. These features are used for classification. Artificial neural network classifier is used for classification. In order to improve the accuracy of classification, the ANN is optimized by Genetic Algorithm.