Melanoma Skin Cancer Detection Based on ABCD Rule

Skin cancer is the most common cancers in the last years, especially in the human body; the Melanoma is the most destructive type of skin lesions. Detect cancer is important at the initial stage, but only an expert dermatologist can detect which one is non-melanoma and melanoma. Computer-aided diagnosis (CADs) application to skin cancer is relatively understudied. The purpose of this paper is the automated detection of Melanoma via digital image processing. In this project, the algorithm consists of automatic ABCD (asymmetry, border irregularity, colour, and dermoscopic structure) rule of dermoscopy lesions images is implemented. Before that, we use hair removal as a pre-processing step which is based on morphological filter and thresholding. Finally, the lesions are classified as either melanoma or benign. The used dataset is containing 200 dermoscopic images, where 120 are benign lesions and 80 malignant melanomas. The proposed method shows an accuracy of 93.2%, 92.59% specificity, and 90.15% sensitivity.