Comparative Study of DNN Models for Skin Cancer Detection

Malignant melanoma, the most dangerous form of skin cancer, has become more common in recent years. The persistent global increase in this cancer, as well as the high medical costs and death rate, have made early detection of this malignancy a priority. Melanoma survival and cure are directly proportional to its thickness; if it can be found early, the survival rate will be enhanced. Despite the fact that significant progress has been achieved in the identification of skin malignancies, there are still significant issues. In comparison to the naked eye, computer-based detection systems can enhance the diagnosis rate of melanoma by 5 – 30 percent. Because visual perception frequently contains flaws, the need for a second viewpoint with greater precision and consistency is underlined. On the other side, it minimizes the tasks and responsibilities that physicians are responsible for. Many studies have been conducted in the area of automated melanoma detection. The potential benefits of such investigations are enormous and unquantifiable. Furthermore, the challenges are numerous, and fresh contributions in this field are greatly valued. However, it is widely known that more reliable and consistent detection methods necessitate a higher level of accuracy. The goal of this paper is to provide a skin cancer diagnosis system that can automatically identify lesions as malignant or benign. Preprocessing, segmentation, feature extraction and selection, and classification are some of the methods used in an automated skin cancer diagnosis.