الفهرس 
-Only 14 pages are availabe for public view
 |  | from | 31 |  |  |
| | | from | 31 | | |
Abstract
As a medical imaging technique, computed tomography (CT) is quite useful for doctors to analyze the pathological changes of the biological organs. In order to reduce deaths, the diseases must be detected accurately in the early stage. The lung cancer is one of the most common internal malignancies also one of the leading death causes. The main problem of lung segmentation from CT images is related to low contrast between lung and nearby organs intensities. Lung sometimes presents in different dimensions and makes the detection and segmentation even more difficult. In this thesis, the original image contains on the dark data so we will use the different stages to enhance the image. Step1, use the contrast stretching of the original image to enhance the image. Step2, remove the noise by different types of filters such as wiener, diffusion, median and Gaussian filter to get the best due to high PSNR and low MSE .In the next stage, image segmentation to isolate the region of interest (ROI) and easier to analyze. This process is done by two steps. Step1, lung segmentation using two methods such as automatic thresholding and active contour method. Step2, tumor segmentation using two methods such as thresholding with morphological operations and watershed method. When we compare between the algorithms due to the active contour the best method because the performance at 96%.In the next stage, feature extraction to extract the features from the above stage such as shape, area by texture features such as Gray Level Co-occurrence Matrix( GLCM). In this stage the features of tumor image at input of the network. We will use the Back Propagation Artificial Neural Network (BPANN) with Leven –Berg Marquardet algorithm to increase the training .Finally, the classification the grades of tumor by calculating tumor burden to grade1,grade2 and grade3.Objectives of thesis :1-Design an algorithm for image enhancement to enhance the low contrast of images.2-An algorithm for filtering the noise that found in the images by using one of the different types of filter (Diffusion filter) is introduced.3-Design an algorithm for image segmentation methods to easier the analyze.4-Design an algorithm for image classification and training our output data to neural network5-Finally, the classification of grades the tumor by tumor burden.