Skin Cancer Detection And Classification in Humans

Abstract

Skin cancer has become one of the major health issue. Skin cancer is of two types. They are malignant melanoma and Benign melanoma. Benign melanoma is not a deadly skin disease to humans where as malignant melanoma is a deadly skin disease to humans. The input to the proposed system is the skin lesion image and then apply median filter to extract the skin lesion image from the healthy skin. Some differenciable factors of malignant(cancerous) melanoma and benign melanoma were extracted using Gray Level Co-occurance Matrix (GLCM) procedure. The extracted features were submitted as input to Artificial Neural Network (ANN) classifier. This classifier classifies the infected skin region and produces the output as Normal skin or Melanoma cancer. The image database contains total number of 90 different dermoscopy lesion images including normal, atypical, and melanoma cases. The tested results shows that the proposed system is efficient and can achieve the classification of the normal, atypical and melanoma images with accuracy of 97.3%, 96.7% and 98.5%, respectively. Hence, the computer based diagnosis(identification of a disease) system can enhance the speed of diagnosis.