الفهرس 
ACKNOWLEDGMENTOnly 14 pages are availabe for public view
 |  | from | 161 |  |  |
| | | from | 161 | | |
Abstract
This study adopts the application in medical education to help medical science students and recently graduated doctors to easily identify the skin cancer disease and its symptoms which in turn helps in detection. Skin cancer is a serious disease that is difficult for students of medical science and early graduated doctors without the experience of recognition because of the overlap of symptoms and similarity with normal skin moles. The images of the skin are sensitive and need for accuracy. For all of these challenges, there is a need to develop an efficient and automatic algorithm to segment skin lesion from skin cancer image. This study presents a proposed algorithm to segment skin cancer images accurately as it is an important step towards skin cancer detection. The proposed algorithm based on a combination of the watershed and edge-based segmentation techniques including six main image processing stages. The first stage is reading the input image from the dataset. The second stage is image preprocessing including several steps which has an important role in enhancing the segmentation stage where the input image converted to the grayscale. Then image smoothing and filtering applied through Gaussian filter. Image contour enhanced by applying morphological closing operator and the process of denoising applied using 2-D wavelet Transformation. After enhancing the image quality, the segmentation stage started. The segmentation process passes through five main steps: mark the background objects, filling holes and reconstruct the lesion border, mark the foreground objects, apply the watershed transformation and apply edge detection. The segmentation process followed by a post-processing stage to ensure that the resulted segmented lesion and its border free of holes and outside noise. The resulted segmented lesion then subjected to feature extraction stage to extract features that help in identifying the lesion and in the classification stage. The proposed algorithm extracts three types of features which are: texture features, shape features and color features. The classification is the last phase which used the features vectors resulted from the feature extraction. The dataset first is divided into a testing and training set and then apply classification using KNN classifier. To evaluate the performance of the proposed algorithm, performance measures include: RMSE, Accuracy, Specificity, Precision, Recall, and F-measure are employed. The proposed algorithm achieved high performance with 97.75% of accuracy. It demonstrated superiority in performance compared to the traditional methods and other segmentation algorithms in related studies.