الفهرس 
AbstractOnly 14 pages are availabe for public view
 |  | from | 130 |  |  |
| | | from | 130 | | |
Abstract
”Nowadays, Skin cancer is one of the most common cancers. However, early and fast detection of skin cancer can save the patient’s life. In this thesis, an alternative method to detect and classify a lesion by using the current technology is developed, to avoid patients the painful and time-consuming process of biopsy method. The skin cancer detection systems are beneficial for humans, since the lesion can be classified within some seconds. The main focus of this thesis is to propose skin cancer (melanoma) detection through introducing developed image processing approach for a computer-aided system. In this thesis, we present two skin cancer detection systems. The two methodologies consist of basic components and different methods are used in each stage. In first proposed system, a neural network (NN) is created, that can differentiate malignant from benign nevus. The NN architecture is analyzed by evaluating it during the training process. Some performance metrics, such as accuracy, specificity, and sensitivity are calculated. Different filters are applied in order to reduce noise and enhance the image, and to extract correctly the necessary features, which are used to train the classifiers. Then automatically the skin lesion is segmented to evaluate the boundary and position of the lesion. Features just like asymmetry, boundary irregularity, color and diameter (ABCD) have been extracted. This method provides the necessary features to categorize a lesion in benign or malignant. Different classification algorithms are trained with the extracted features to fit the input data. The most benefits from the first proposed technique were obtained, which achieved in an accuracy of approximately 98 %, sensitivity of approximately 100% and specificity of approximately 96.2%, which is effective in detecting the presence or absence of melanoma. The second proposed system includes four main stages: extracting RoI, data augmentation, deep feature extraction, and classification, with using four methods based on convolutional neural networks (CNNs). where GoogleNet, ResNet-50, AlexNet, and VGG19 were investigated and compared. Transfer learning technique was applied for each deep learning model. The second proposed technique showed the ability to improve the classification accuracy with an accuracy of 99.31% for ISIC database and 100% for CPTAC-CM database. The results and experiments showed the privilege over the state-of-art techniques of early skin cancer detection systems, which applied on the same dataset. The findings obtained suggest the viability of the proposed method of diagnosis of skin cancer. Using this technique, encouraging results are obtained, these results proved the promise of the proposed systems for grading skin cancer. The thesis comprises five chapters that can be summarized as follows: Chapter (1) gives an overview of the study conducted. It has comprised of the introduction of skin cancer detection system which gives a brief idea on what is a skin cancer and diagnosis is about. It also explains briefly the problem statement, objectives, the scope of the study and proposed system. Chapter (2) gives an overview of human skin biology, skin cell structure, layers of skin, melanocyte, physical properties of human skin and skin cancer and its types. It also explains previous research work related to this thesis. Chapter (3) describes the procedure of this study and reveals the techniques and the algorithms that will be used in performing the research. It contains information about the datasets used and how the CNN was trained, validated and tested. The best method and techniques used for this system is described in detail. One of the software processes is chosen as methodology. Implementation of process that is involved during development of this system is explained in detail. Describes two standard skin cancer datasets, two skin cancer detection systems, with a detailed description of each system framework. Chapter (4) explains the results that obtained for both proposed skin cancer detection systems, and includes comparison with relevant research findings. Chapter (5) presents the findings and the most important directions for future work.