Detection techniques for melanoma diagnosis: A performance evaluation

Melanoma which is most commonly spread skin cancer in United States, UK and Australia and it has been estimated that about 12,650 people died from this cancer in the year 2013 in UK. Thus, early detection and diagnosis of the tumor is most important. And according to survey it was found that 60% of claim lodged against Medical Protection Society for incorrect diagnosis by doctors or negligence of medical reports due to system failure. So, accurate diagnosis is required for effective treatment of melanoma. This paper presents various diagnostic techniques such as Menzies scale method, Seven Point checklist, Asymmetry, Border, Color, Diameter (ABCD)rule based method and Pattern Analysis method for early diagnosis of cancer. The methodology which is been followed in diagnosis of skin cancer is also presented and a comparative study of various edge detection, an image preprocessing and segmentation technique are proposed. In our study, analysis has been done on number of samples melanoma clinical images and it has been observed that Canny edge detection and preprocessing using Otsu method is the best approach. Due to importance of image segmentation a number of algorithm have been proposed and based on image that is inputted in the algorithm should be chosen to give the best results.