A Classification of Skin Lesion using Fractional Poisson for Texture Feature Extraction

Melanoma has been widely recognized as a dangerous type of skin cancer due to its ability to rapidly spread through the human body. An early detection of melanoma has proven to improve the success rate of clinical treatment. This problem may further extend to the field of computerized melanoma detections. In this study, we propose a method using texture features based on fractional Poisson to classify melanoma. Eight masks were created for eight direction. Mean and standard deviation were computed for each direction to create feature vector from images. Three classification models: Support Vector Machine (SVM), AdaBoost-M1, and K-nearest neighbors algorithm (K-NN) were used for separating melanoma from other types of skin lesions. It was found that AdaBoost-M1 provided the highest 100% results for the correct classification rate, specificity and sensitivity.