科研之友
微信
新浪微博
Facebook
分享链接
摘要
Breast cancer is one of the most dangerous tumors for middle-aged and older women in China, and CR mammography is its most reliable early detection method in the clinic. In order to assist the radiologists in reading the CR mammograms, a novel computer-aided detection (CAD) system was presented in this paper. At first, it used a segmentation threshold optimization method for the pectoral muscle by applying a series of regions of interest (ROIs) with various sizes, presented a zonal Hough transform to segment the rough region, and then applied polygon fitting to approach the edge of the pectoral muscle. Secondly, it used a thresholding with hysteresis in the wavelet domain to locate the microcalcifications (MCs), and applied an ANFIS-based filling dilation algorithm to extract the MCs. Then, it utilized a model-based detection algorithm for the masses, using different information extraction ways to detect different kinds of masses, and applied a filling dilation restricted with Canny edge detector and energy field to segment the masses. Finally, three neural networks were used to classify the lesions for comparison, and MLP was confirmed as the most stable one. And the proposed CAD system had a totally better performance than the conventional methods in the experiments. ? 2010 Binary Information Press.
- 出版日期2010
- 单位浙江大学
