Although melanoma is the least common skin cancer, it is the most dangerous type of skin cancer. On the other hand, if it is diagnosed in the early stages, the chance of success in its treatment is quite high. Dermoscopy is a non-invasive imaging technique for diagnosis. Considerable attention is currently paid to the development of computer-aided diagnostic systems to assist in the clinical evaluation of dermatologists. Image segmentation is the process of dividing an image into meaningful parts, each of which contains different features. Segmentation also helps to find the region of interest in a particular image. The main goal is to make the image simpler and more expressive. In this study, dermoscopy images from Hospital Pedro Hispano (HPH) were used. In order to reduce the noise, median filtering was applied to the images as a preliminary image processing step. Melanoma skin cancer images were segmented by Otsu thresholding method, which is the most preferred and used method among thresholding methods. Segmentation performance values on the images are shown simultaneously in the prepared MATLAB GUI interface. According to the results of the performance analysis, for OTSU, accuracy 96%, sensitivity 88%, specificity 99% and precision 99% were calculated. In conclusion, it has been shown that the detection accuracy of this segmentation method for melanoma skin cancer is good.
Anahtar Kelimeler: Image segmentation, Filtering, Thresholding method, Dermoscopy
|
