3-D Fuzzy Tumor Detection for Whole Breast Ultrasound Image

• Main Authors: Tzu-Hsuan Chen, 陳子軒
• Other Authors: Ruey-Feng Chang
• Format: Others
• Language: en_US
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/47330193033563817602

碩士 === 國立臺灣大學 === 資訊工程學研究所 === 96 === The breast cancer is the most frequently cancer in women and it is the second rank for the death caused by cancer after cancer of lung. The earlier detection and improved treatment are effective to reduce deaths due to breast cancer. Breast ultrasound is a quite important complementary imaging modality with mammography to detect breast cancer early. However, it needs a lot of physician time to screen a patient by the manual ultrasound. Recently, the automatic whole breast ultrasound has been developed to save the physician for screening the breast. Because a lot of images are obtained for a case, the physician still takes a lot of time to diagnosis. Hence, a computer-aided tumor detection system is proposed to find suspicious regions of tumors for saving the diagnosis time. In this system, the image is firstly pre-processed by removing black regions and sub-sampling to reduce the detecting time and the sigmoid filter to enhance the boundary between tumor and normal tissue. Then, a three-dimensional (3-D) fuzzy technique is adopted to detect tumor regions in the breast. This method classifies the image as three categories, tumor, boundary, and normal tissue according to the intensity and edge information of image. The detected tumor regions are not all real tumor regions. Some of them may be darker regions, such like shadow or nipple. In order to obtain actual tumor regions, the connected component labeling groups each voxel of tumor regions individually and some tumor criteria are proposed to filter out non-tumor regions according to the characteristic of these connected components. In the experiment, 45 test cases are tested by the proposed tumor detection system. By experimental results, almost all tumors can be found by this system and the sensitivity is up to 85.9% (55/64) with 1.71 false-positive rate per case. This means that the proposed system satisfies the high detecting performance with low false-positive rate and is a good tool to help the diagnosis of doctors.