作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2013, Vol. 39 ›› Issue (5): 257-260. doi: 10.3969/j.issn.1000-3428.2013.05.056

• 图形图像处理 • 上一篇    下一篇

基于稳健特征统计的医学影像分割算法

刘玲玲1,康晓东1,王 昊1,2,耿佳佳1   

  1. (1. 天津医科大学医学影像学院,天津 300070;2. 河北大学附属医院,河北 保定 071000)
  • 收稿日期:2012-06-04 出版日期:2013-05-15 发布日期:2013-05-14
  • 作者简介:刘玲玲(1989-),女,硕士研究生,主研方向:图像处理;康晓东,教授、博士;王 昊,工程师、硕士研究生;耿佳佳,硕士研究生
  • 基金资助:
    国家自然科学基金资助项目(60603027);天津市应用基础与前沿技术研究计划基金资助项目(05YFJMJC11700)

Medical Image Segmentation Algorithm Based on Robust Feature Statistics

LIU Ling-ling 1, KANG Xiao-dong 1, WANG Hao 1,2, GENG Jia-jia 1   

  1. (1. School of Medical Imaging, Tianjin Medical University, Tianjin 300070, China; 2. Affiliated Hospital of Hebei University, Baoding 071000, China)
  • Received:2012-06-04 Online:2013-05-15 Published:2013-05-14

摘要: 提出一种基于稳健特征统计的医学影像分割算法。由用户提供标记的种子点,通过稳健统计量描述种子点及其周围点对象的特征,使得分割的边缘更光滑,且对噪声不敏感,对边缘进行轮廓演变,基于稀疏场方法完成曲线演化,找到理想边界。实验结果表明,在MR腹部肝脏分割中,该算法5种评价指标的最终得分为73分,高于区域增长算法和快速水平集算法,肝脏分割时间为123 s,能较好地分割MR和CT图像中的器官和肿瘤。

关键词: 图像分割, 稳健特征统计, 特征向量, 稀疏场方法, 水平集, 曲线演化

Abstract: A medical image segmentation algorithm based on robust feature statistics is proposed in this paper. Users provide seeds, and this paper uses robust feature statistics to describe the features of objects, makes contour smoothly and decreases the effect of noise. It makes contour evolution fast and according to the Sparse Field Method(SFM) to get the outline of objects. Experimental results show that this algorithm gets 73 points in 5 kinds of evaluation indicators for MR abdominal liver segmentation, is better than the region growing algorithm and rapid level set algorithm, liver segmentation time is 123 s, and can make good segmentation of organs and tumors in MR and CT images.

Key words: image segmentation, robust feature statistics, feature vector, Sparse Field Method(SFM), level set, curve evolution

中图分类号: