الفهرس 
Acknowledgments
The Modified Perceptron Algorithm for Breast Cancer DiagnosisOnly 14 pages are availabe for public view
 |  | from | 86 |  |  |
| | | from | 86 | | |
Abstract
Thesis Summary
Breast cancer is the most common cancer in women worldwide. It is also the principle cause of death from cancer among women globally. The cancer rates are high in more developed countries, whereas, rates in less developed countries are low but lately increasing. Early detection of this disease can greatly enhance the chances of long-term survival of breast cancer victims. There are many medical techniques for early detection of breast cancer. In this thesis, data from two different medical techniques: fine needle aspiration (FNA) biopsy and magnetic resonance imaging (MRI) images to classify by artificial neural network (ANN). This thesis presents a new ANN algorithm named as modified perceptron algorithm for data classification (MPADC) formed to classify a nonlinearly separable data. If the classification problem is not linearly separable then the original perceptron algorithm will be executed infinite number of times. The original perceptron algorithm modified by adding a new termination criterion using evaluation vector. The evaluation vector used to monitor the accuracy and thence setting predefine termination accuracy interval to avoid endless execution times. We apply MPADC to the Wisconsin breast cancer original dataset (WBCD) that consist of visually assessed nuclear features of FNA taken from patients’ breasts. The average classification accuracy of MPADC algorithm appears to be very promising. In MRI, the second medical techniques, tumor tends to be darker and more intensity than the surrounded tissues. because when the tumor grows it needs to support its center with blood to still alive, so
the mean values of the ratios of tumor to normal blood flow and blood volume are significantly higher than those for benign or normal tissue. The MRI is one of the methods that can detect blood flow and blood volume. In this thesis we use scale invariant feature transform (SIFT) feature extraction algorithm that care with darkness and intensity of the images to extract features of MRI breast cancer images. This hybrid between SIFT and our algorithm called breast MRI cancer classifier (BMRICC). BMRICC compared with different ANN algorithms. The result shows that BMRICC is better than many ANN classification algorithms.
Supervisors Signature Student Signature