基于稳健特征统计的医学影像分割算法  

Medical Image Segmentation Algorithm Based on Robust Feature Statistics

作　　者：刘玲玲[1] 康晓东[1] 王昊[1,2] 耿佳佳[1] 

机构地区：[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

出　　处：《计算机工程》2013年第5期257-260,共4页Computer Engineering

基　　金：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)

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。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.

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 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]