Özet:

Melanoma is one of the most fatal types of skin cancer caused by structural DNA damage, but the survival rate is quite high in case of early diagnosis. The diagnosis of melanoma is made by the visual examination of the dermoscopic images of the injury area by a dermatologist. The exact diagnosis is taken with a tissue sample from the injury area and appears after a pathological examination. The pathological examination is both a long-term and expensive method. A large number of computer-backed diagnostic systems have been developed to minimize the errors that dermatologists can make and help the doctor to correct diagnosis of melanoma. It is critical because the proper division of the lesion area directly affects the output of characteristics and classification performance. In this study, a series of pre-dimension steps were first applied, such as re-dimension, black frame removal and hair removal. The traditional Ossu, Ossu-based HA, SA and PSO methods were then applied to find the appropriate marginal value to separate the injury area from healthy skin by obtaining the histogram of the grey-level image. According to the most suitable marginal value found, the image is converted into a binary image and the final damage area is obtained. As a result, when the working time is calculated according to the average number of iterations, it was observed that the PSO, HA, SA methods found an optimal marginal value faster, respectively, by 40%, 29% and 28.5%, compared to the traditional Ossu method, without giving rise to the success of dividing.

Anahtar Kelimeler: