基于模糊连接度的FCM分割方法在医学图像分析中的应用  

APPLICATION OF A NEW MEDICAL IMAGE SEGMENTATION METHOD BASED ON FUZZY CONNECTEDNESS AND FCM

作　　者：林瑶[1] 田捷[1] 张晓鹏[1] 

机构地区：[1]中科院自动化所人工智能实验室,北京100080

出　　处：《中国体视学与图像分析》2001年第2期103-108,共6页Chinese Journal of Stereology and Image Analysis

基　　金：国家自然科学基金 (No.6 99310 10 ;6 0 0 710 0 2 );86 3计划基金(86 3 -30 6 -ZD -11-3)资助

摘　　要：图像分割的一个重要应用领域是医学图像的分割。我们针对医学图像的模糊特点和实际应用的要求 ,结合模糊连接度阈值分割和模糊C均值聚类两种分割方法的优点 ,提出一种新的交互式医学图像分割方法。首先计算整幅图像的模糊连接度 ,通过阈值分割提取出感兴趣的对象 ,并将模糊连接度作为图像的冗余特征 ;然后在由冗余特征和原图像特征构成的二维聚类空间中 ,利用模糊C -均值聚类方法优化上一步骤的分割结果 ,提高分割准确度。我们以CT和MR图像为实验对象进行了验证 ,实验结果表明这是一个有效的方法。Medical image segmentation is one of the important application areas of image segmentation. Considering the fuzzy character of medical images and requirement of practicability, we present a two-stage interactive image segmentation method. First, the object with interest is extracted and a new image scene is constructed using fuzzy connectedness-based method. The new scene together with the original scene forms a two-feature space. Second, the fuzzy C-means method is adopted on the clustering space in order to improve the accuracy of segmentation. We used the method to segment some CT and MT images, the results show that this method deserves our further research.

关 键 词：医学图像分割 FCM 模糊连接度 聚类分析 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]