Improved classification of skin lesions using shape and color features

It is very important to scan the skin lesions in general population, and to forward suspicious cases to the dermatologists. Typically, the number of dermatologists is not enough for such a wide scanning campaign. Hence, an automated system and algorithms are required that can detect suspicious skin lesions. These systems should be highly accurate and sensitive. In this work, an automated system is developed to detect suspicious skin lesions from digital images using shape and color features. This system is tested on images in the PH2 database. The developed system has 91.5% accuracy, 92.5% sensitivity and 87.5% specificity.