一种基于水平集多尺度乳腺肿块分割方法.pdfVIP

31 6 Vol31 No6 2010 6 Ch inese Journal of Scientific Instrum ent Jun. 20 10 * 1 1 1 1 2 2 虞红伟 , 厉力华 , 徐伟栋 , 刘 伟 , 张 娟 , 邵国良 ( 1 310018; 2 310022) : , , C-V , ; C-V , , , , C-V RSF 89, 0. 236 10. 300 4 0. 373 8, : ; ; ; X : TP391 : A : 520. 6040 M ulti-scale approach for breastm ass segm entation using level setm ethod 1 1 1 1 2 2 Yu Hongw ei, Li Lihua , Xu W eidong , LiuW ei, Zhang Juan , Shao Guoliang (1 Institutef orB iomed icalEng ineering and Instrumentation, HangzhouD ianzi University, Hangzhou 310018, hina; 2 D ep artment of Radiology, Zhejiang ancerH osp ital, Hangzhou310022, hina) Abstract: A s an mi portant step of breastm ass detection, mass segmentation p lays an mi portant role in many comput- er-aided d iagnosis ( CAD) system s for mammography. In th is paper, w e propose a no el schem e for mass segmenta- tion in mammography, wh ich is based on le el setm ethod and mu lti-scale analysis. M ammography is firstly decom- posed into a sequence of mi ages from fine to coarse using Gaussian pyram id, the C-V model is then applied at coarse scale, and the obtained rough contour is used as the initial contour for segmentation at fine scale. A local acti e con- tour model based on the mi age local information is utilized to refine the rough contour locally at fine scale and pre ent the defect of the C-V modelw hen it is applied on the mi ages w ith asymmetr ic gray distribution. In order to mi pro e the adapti ity of our method, the features of area and gray le el extrac