الفهرس 
AbstractOnly 14 pages are availabe for public view
 |  | from | 131 |  |  |
| | | from | 131 | | |
Abstract
Computer Aided Detection (CAD) and Computer Aided Diagnosis (CAD/x) became a promising field in the area of applications of medical images. Kidney carcinoma is a fetal disease, and early detecting is an important factor for surviving. Computed Tomography (CT) is the most efficient technique to detect the organs, CT scan provide a very accurate anatomical structure, but CT must be with a low dose (as precautions procedures), so a low contrast is the result, which renders the interpretation or distinguishing the organs, so hard. In this thesis an algorithm had been designed to enhance, segment and refine the lower torso (CT) datasets, by aid of (CAD-CAD/x), two kidneys were detected from the abdominal image, the tumor (if it was) had been segmented from the kidney. Analytical study had been addressed to determine the tumor features (area-location-grade). The data saved in a predetermine locations, to generate a complete diagnosis message or to develop the algorithm. On the other hand, a Grey Level Co-occurrence Matrix (GLCM) and Artificial Neural Network (ANN) had been implemented for classifying the tumor nature. In addition, the algorithm was converted into an integrated program, physicians can follow the algorithm recommendations or rely on their own vision, to write the medical report. This thesis had been done, by aid of theories, functions, and applications of digital image processing, MATLAB , ANN, and MS EXCEL. The study is organized according to the following eight chapters: Chapter 1 is a general introduction to computer-aided detection(CAD), its applications and a block diagram of the proposed algorithm, then a complete flowchart of the algorithm with a brief explanation of its blocks, eventually the thesis organization Chapter 2 explaines the description, anatomy, and functions of the kidney and nephrons, a brief explanation of kidney carcinoma, then the common methods for kidney cancer detection by imaging. Chapter 3 a brief explanations for CT- elements, different generations, the operations theory, and CT strengths. Eventually a brief explanation for images reconstruction technique. Chapter 4 the first stage of the tumor detection algorithm, which is a preprocessing stage. In this chapter we explain the theoretical meaning of image enhancement techniques, in digital image processing. The first step is the normalization, then proposed a contrast stretching technique, which analyze the image to yield a good result, by using Graphical User Interface (GUI), then we explains the theories of the most common noise in CT medical images, by the aid of equations. Noise removal techniques will take place, a comparative study between four filters is implemented, by the aid of quality metrics Mean Square Error(MSE), Peak Signal to Noise Ratio (PSNR) and Human Visual System(HVS), eventually the convenient filter is selected. Chapter 5 begins with the flowchart of the segmentation stage, which consists of two steps, then a theoretical explanation of segmentation in digital image processing and then full explanation for the first step, which is a comparative study between three segmentation techniques(spine detection-active contour-thresholding), to detect the two kidneys. Then the second step, which is the abnormality detection step, it is implemented by aid of comparative study, between three techniques (Otsu-Watershed-Thresholding), then their results, The weighting was postponed to the next chapter. Chapter 6 consists of four sections. first section is, the morphological operations stage, which consists of six consecutive steps, the main idea is clarifying the tumors, to decide the tumor existence then a proposed technique is selected by aid of ‘gold standard rule’ and the metrics (Sensitivity, Specificity & Accuracy). Second section is the tumors classifications, which is implemented by Grey Level Concurrence Matrix (GLCM) and Artificial Neural Networks (ANN), to decide the tumor nature. Third section is collecting the slice data, storing it in MS Excel sheet, then exchange the data with MATLAB space, to create the kidney complete diagnosis messages, as a CAD/x system. Last section was a complete statistics for thesis techniques, include tables and graphs, then a comparative study is addressed, between thesis evaluations and another three recent techniques. Chapter 7 all used techniques had been converted into a complete program, with menu options, the chapter begins with the main menu and all of the algorithm results, then every option with its visual results. Chapter 8 is a thesis conclusion. Fifty-six CT images for seven patients consisted the thesis data set. Overall, algorithm evaluation were Sensitivity (45-50) (90%) ,Specificity (6-6) (100%) and Accuracy (51-56) (91%). Also, in ANN, to classify the tumors, fifty images were tested from six patients, evaluations were (25 benign, 25 malignant). ANN achieved the following evaluations: SPECIFICITY (25-25) (100%), SENSITIVITY (24-25) (96%) which mean that the ACCURACY got (49-50) (98%).