Melanoma Recognition System

Malignant melanoma is nowadays one of the leading cancers among many white-skinned populations around the world. Change in recreational behavior together with the increase in ultraviolet radiation cause a dramatic increase in the number of melanomas diagnosed. The rise in incidence was first noticed in the United States in 1930, where one person out of 100 000 per year suffered from skin cancer. This rate increased in the middle of the eighties to six per 100 000 and to 13 per 100 000 in 1991. The numbers are also comparable to the incidence rates observed in Europe. In 1995, in Austria, the incidence of melanoma was about 12 per 100 000, which reflected an increase of 51.8 % in the previous ten years, and the incidence of melanoma shows a still increasing tendency. But on the other hand, investigations have shown that the durability of skin cancer is nearly 100%, if it is recognized early enough and treated surgically. Whereas the mortality rate caused by melanomas in the early sixties was about 70 %, nowa survival rate of 70% is achieved, which is mainly the result of early recognition. Because of the higher incidence of malignant melanoma, researchers are more concerned more and more with the automated diagnosis of skin lesions. Many publications report on isolated efforts in the direction of automated melanoma recognition by image processing. Complete integrated dermatological image analysis systems are hardly found in clinical use or are not tested on a significant number of real-life samples. We have developed a fast and reliable system that is capable to detect and classify skin lesions with high accuracy. We use color images of skin lesions, image processing techniques, and AdaBoost classifier to distinguish melanoma from benign pigmented lesions. As the first step of the data set analysis, a preprocessing sequence is implemented to remove noise and undesired structures from the color image. Second, an automated segmentation approach localizes suspicious lesion regions by region growing after a preliminary step based on adaptive color segmentation. Then, we rely on quantitative image analysis to measure a series of candidate attributes hoped to contain enough information to differentiate melanomas from benign lesions. At last, the selected features are supplied to the AdaBoost algorithm to build a strong classifier.