الفهرس 
INTRODUCTIONOnly 14 pages are availabe for public view
 |  | from | 95 |  |  |
| | | from | 95 | | |
Abstract
Breast cancer becomes a significant public health problem in the world. It is a challenging task to classify accurately the benign-malignant mammography patterns. This work proposes a new fully automated computer-aided diagnosis (CAD) system for breast cancer diagnosis as malignant or benign with high accuracy and low computational requirements. The Expectation-Maximization (EM) algorithm is proposed to detect and extract automatically the Region of Interests (ROIs) within mammograms. The standard textural, shape, and statistical features of ROI are extracted and combined with multi-resolution and multi-orientation features derived from a new hybrid feature extraction technique using a Wavelet-Based Contourlet Transform (WBCT), allowing for accuracy improvement over other standard approaches. A hybrid feature selection approach based on combining the Support Vector Machine Recursive Feature Elimination (SVM-RFE) with Correlation Bias Reduction (CBR) technique is proposed. Also, for the first time, a new similarity–based learning algorithm called Q-classifier for benign-malignant classification is proposed. The proposed CAD system is applied to real clinical mammograms, and the experimental results demonstrate the superior performance of the proposed CAD system over other existing CAD systems in terms of accuracy 98.50 %, sensitivity 98.99 %, specificity 98.02 %, and computational time 2.2 sec achieved by the proposed CAD system. This reveals the effectiveness of the proposed CAD system in improving the accuracy of breast cancer diagnosis through real time diagnosis systems. In order to evaluate the efficiency of this implemented CAD systems, it was compared with another combination of high advanced techniques. These techniques included the Watershed technique for segmentation, a hybrid feature selection approach in which the Genetic Algorithm (GA) and the Support Vector Machine (SVM) are combined along with the Mutual Information (MI), and finally, the kernel SVM classifier is included in the comparison. The results show that best results are derived from the following combinations: WBCT + SVM-RFE-CBR + Q-classifier, achieving an accuracy of 98.5% for benign-malignant tumors classification.