三维超声心脏图像的模糊聚类分割  

Fuzzy Clustering Segmentation for 3D Echocardiographic Images

作　　者：钟一民[1] 林其忠[1] 余建国[1] 薛海虹[2] 孙锟[3] 

机构地区：[1]复旦大学电子工程系,上海200433 [2]新华医院,上海200092 [3]上海儿童医学中心,上海200127

出　　处：《上海生物医学工程》2007年第2期72-75,共4页Shanghai Journal of Biomedical Engineering

基　　金：上海市科委资助项目(05449611)

摘　　要：采用三维超声心动图对小儿先天性心脏病进行诊断与治疗能达到比传统二维超声心动图更直观的效果。然而由于超声图像质量较差,三维超声心动图的可视化效果往往无法达到医生的要求。本文对三维超声心脏图像进行分割,以改进超声图像的可视化效果,并为参数提取等提供基础。首先采用快速的模糊c均值聚类得到初始分割结果;然后利用图像多分辨率技术进行修正;接着结合图像的对比度进行进一步的分割;最后,把处理后的图像用绘制的方法显示出来。本文的结果对超声图像的可视化效果有一定的改善.The 3D echocardiography ultrasound assessment is more intuitionistic than the 2D one in the domain of diagnosis and treatment of children＇ s congenital heart diseases . Because of the poor quality of ultrasound images, the visualization results are always under surgeons＇ requirements. Here a segmentation technique for 3D echocardiography is proposed to enhance the visualization quality . First we use a fast fuzzy c - means algorithm to get the initial segmentation result . Then a multiscale technique is used. After that, a texture feature is integrated into the algorithm. Finally, we visualize the images by rendering. The results show some improvements in the visualization of 3d ultrasound images.

关 键 词：图像分割 三维超声 模糊聚类 多分辨率 纹理对比度 

分 类 号：R318[医药卫生—生物医学工程]