Microcalcifications Detection in Digital Mammograms
| Autor: | Rou-yi Lin, 林柔伊 |
|---|---|
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 95 Breast cancer is the top of woman’s malignant neoplasm. Many researchers have investigated methods for improving the screening accuracy. Early finding and treatment could reduce the mortality rate of breast cancer. The common examinations of breast cancer include ultrasound, mammography, MRI, etc. Mammography is more effective and cheaper for routine examination than the others. In clinic applications, to check the early stage of the breast cancer is useful for the prognosis. One of the important early symptoms of breast cancer in mammograms is the appearance of microcalcification clusters. This study aims to locate microcalcifications automatically. Each set of mammographic images consists of two views: a cranio-caudal view (CC) and a medio-lateral view (MLO) in both left and right breasts. Thus each examination generates four films when a patient accepts the breast cancer screening. In the experiments, we used about 30 cases, including 116 films from Buddhist Tzu Chi General Hospital. The images are separated into three groups: normal type (71), mass (14), and microcalcifications (41). We applied top-hat transform to enhance the images for easily locating microcalcifications. After thresholding, connected components were found. Then, we generated a feature set including shape and textural features for classification of the components. In this study, we use the support vectors machine (SVM) neural network as a classifier. The final correct rate from the support vectors machine is 98.4%. The SVM has good performance for the classification of microcalcifications. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|