Skin Lesion Classification: An Optimized Framework of Optimal Color Features Selection

Melanoma is the most common and deadly kind of malignancy among all the existing types of cancers, worldwide. Globally, the incidence rate of melanoma rising in recent decades. Responses on a survey, in USA about 192,310 new cases are diagnosed while 7,230 deaths have been occurred due to melanoma in 2019. This ratio can be decreased if it is detected at an early stage. A novel systematic approach for skin cancer detection based on optimal feature selection is proposed in this work. In the normalization step, it differentiates the lesion region from the surrounding skin region by using a linear contrast stretching technique. Later, various type features are computed and put to optimal feature selection approach name higher entropy value features (HEVF). Optimized and best features are selected and classified using SVM classifier and evaluated on ISBI 2017 dataset. As a result, the proposed systems get a performance of 96.2% which is improved as compared to existing techniques.