基于纹理分析和双边滤波方法对加标记心脏核磁共振图像分割

背景：加标记心脏核磁共振成像方式提供了左心室内外心膜的边缘信息，该边缘信息可由分割图像得到。但是，所引入的标记线加大了这类图像边界分割的困难。 目的：针对目前在加标记心脏核磁共振图像中对左心室分割困难的问题，提出了一种新的自动分割的方法。 方法：首先，使用全局直方图规定化方法增强标记和非标记区域的对比度；然后，利用一种简单的纹理分析方法区分血流充盈的心腔(非纹理)区域和加标记心肌(纹理)区域；再应用双边滤波在保持边界的同时滤掉图像的伪影；最后，用GVF-snake模型自动提取左心室图像的边界。 结果与结论：提出了一种简单的纹理分析方法来移除标记线:用局部窗口中的最大灰度值与最小灰度值之差来代替原象素点灰度值，再运用双边滤波滤除图像伪影并保持边界，最后应用GVF-snake模型实现了左心室边界的有效提取。实验结果显示，该方法能够较好地提取部分加标记心脏核磁共振图像中血流充盈区的边界。 关键词：纹理分析；双边滤波；标记线；左心室；核磁共振图像；数字化医学

Tagged Cardiac Magnetic Resonance Image Segmentation Based on Texture Anlysis and Bilateral Filtering

BACKGROUND: Tagged cardiac magnetic resonance imaging provides some information of cardiac boundary, which can be obtained by segmentation. However, the added tagging lines bring great difficult to accurate segmentation of the cardiac boundaries. OBJECTIVE: Concerning the problem of segmentation of the left ventricle in tagged cardiac magnetic resonance (MR) images, a new method of automatic segmentation was proposed. METHODS: Global histogram specification was firstly applied to improve contrast between tagged lines and untagged region. Afterwards, a simple method of texture analysis was adopted to discriminate ventricle blood-filled (untagged) and tagged (texture) regions. Bilateral filter was applied to filter off some block-like artifacts as well as to preserve the boundary of the left ventricle. Finally, GVF-snake model was utilized for automatically extracting the boundary of the left ventricle. RESULTS AND CONCLUSION: A simple method based on texture analysis was proposed to remove the tagged lines: gray difference computed by subtracting the minimal from the maximal intensities within a local small window replaces the original gray value of a pixel; afterwards, filter off some block-like artifacts as well as preserve the boundary of the left ventricle by using bilateral filter; finally, the boundary of the left ventricle was extracted automatically based on GVF-snake model. Experimental results show that the proposed method can effectively extract the boundary of the left ventricle from tagged cardiac MR images. Li ZL, Yang XM. Tagged cardiac magnetic resonance image segmentation based on texture analysis and bilateral