الفهرس 
Chapter 1Only 14 pages are availabe for public view
 |  | from | 123 |  |  |
| | | from | 123 | | |
Abstract
Several research studies are available in the literature, and concerned with image processing and machine learning techniques for detecting the subtypes of acute lymphatic leukemia (ALL). However, many of these techniques have some disadvantages regarding the detection accuracy and consumed time. Therefore, the present thesis aims to present four improved image processing and machine learning techniques for accurate detection of the ALL subtypes. The first technique is based on image processing procedures which depend on the all parameters corresponding to the known specifications of ALL cells. In this technique, three complementary steps are performed. In the first one, a color segmentation procedure is used to obtain images including only the white blood cell (WBC). In the second step, the histogram equalization and linear contrast stretching procedures are utilized to obtain images for the nucleus. In the third step, images for the cytoplasm only are reconstructed, then, the vacuoles in them may be easily detected. The second technique is an improved classification system based on simple image feature map to identify automatically the ALL subtypes. An adaptive segmentation procedure is performed on peripheral blood smear images to extract the main features from the segmented images of WBCs, nucleus and cytoplasm. A comprehensive study is made on all the possible permutation cases of the features using powerful classifiers which are K-nearest Neighbor (KNN) at different metric functions, support vector machine (SVM) with different kernels and artificial neural network (ANN). The third technique deals with improved a deep learning pre-trained Alex convolution neural network for automatically differentiating between the normal cells and the three ALL subtypes L1, L2, and L3. The used dataset includes 770 blood smear WBC images for normal and abnormal cells. This technique has the ability to extract the image features directly from the input image data without the need for the feature manual extraction saving the time and effort. The fourth technique is a CNN network constructed from a combination between Alex CNN network and a support vector machine (SVM). This network has the advantages of the both networks regarding the automatic feature extraction and the classification accuracy.