摘要
超声图像自动分割技术具有重要的应用价值,同时面临很大挑战。本文针对乳腺超声图像局部分割需求,通过区域生长法自动快速提取初始分割区域作为水平集的初始条件,实现先验区域约束的作用,显著提高分割准确度;基于CV模型进行全局信息和区域信息的拟合,提高了弱边界的定位准确度;增加速度能量项,使零水平集更快地收敛在边界处;对经典水平集模型的一些能量项进行删减调整,以降低计算复杂度。分割实验结果表明,本文方法能够较为准确快速地实现病灶分割,有一定的临床辅助价值。
Accurate and automatic lesion segmentation technology for ultrasound images has important ap- plication value, but also faces great challenges. Aiming at the specified local segmentation demand of breast ultrasound images, initial area was extracted by region growth as priori area constraint for level set Global energy fitting with region energy to locate blur edge was based on CV model. Aspeed energy func- tion was added to accelerate the revolution process of the level set, and some other modifications to degrade the calculation complexity. Experiments showed that the proposed method can extract lesion edges satisfactorily,
出处
《中国体视学与图像分析》
2013年第4期326-335,共10页
Chinese Journal of Stereology and Image Analysis
基金
广东省科技计划项目(2010B060300001)
广东省产学研资助项目(2011B090400037)
关键词
水平集方法
先验区域约束
乳腺超声图像
病灶分割
and play an assisting role in clinic works breast ultrasound image
level set method
prior area constraint
lesion segmentation
