Computer-Aided System for BI-RADS of Magnetic Resonance Breast Images

• Main Authors: Wang, Cheng-Yan, 王程彥
• Other Authors: Yang, Sheng- Chih
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/46327289861859062315

碩士 === 國立勤益科技大學 === 資訊工程系 === 102 === Breast cancer is the major cause of death of women in Taiwan. According to the results of a survey conducted by the Department of Health, Executive Yuan, breast cancer is ranked the first cause of mortality among women aged between 25 and 44; therefore, this disease is a crucial concern for Taiwanese women. Statistics from the American Cancer Society show that women are highly prone to breast cancer after age 40, suggesting the necessity for preventive measures and early detection of the disease. A set of international standards, known as breast imaging reporting and data system (BI-RADS), has been established for MRI-based breast diagnosis, which stipulates standardized procedures for medical history reporting. BI-RADS involves determining the proportion of fiber glands within the breast tissue to the entire breast, and diagnosing whether the patient under examination is at risk of breast cancer. For patients diagnosed with breast cancer, the shape and edge of their tumors must be classified, and whether the tumors are benign or malignant must be specified in the diagnostic report. The purpose of this study was to establish an aiding system for mammography history reporting, which physicians may use as a reference when writing BI-RADS medical reports. Regarding the calculation of the proportion of fiber glands in breast tissue, the fuzzy c-mean algorithm is employed for analysis. Then, regarding tumor segmentation, tumors are processed using the constrained energy minimization method to enhance the brightness, followed by tumor segmentation using the floating support-pixel correlation statistical method developed in this study. Finally, based on the extracted tumor region of interest, the features of the tumor shape and edge are calculated and classified according to BI-RADS standards. The results showed that the shape and edge classification reaches an accuracy of 93%, confirming the feasibility and contribution of the mammography medical reporting system developed in this study.