Abstract: Magnetic resonance angiography (MRA) is the primary choice for vascular imaging in clinic. The precision of vessel segmentation is the key of angiography 3D visualization, diagnosis, treatment and evaluation of vascular disease. By studying the characteristics of the cerebral vascular anatomical structure and MRA images, this paper proposes a cerebrovascular segmentation method based on local Otsu. At first, the original image is preprocessed with anisotropic diffusion filtering. Then, the optimal size of the sub-image is determined by calculating the change rate of the image internal standard deviation between two images of different sub-block sizes. The image is divided into some sub-images of this optimal size. Next, all the sub-images are segmented using the Otsu algorithm and merged to the final segmentation result. The segmentation results are reconstructed by visualization toolkit (VTK) and projected to two-dimensional images for quantitative evaluation. The experimental results show that the method can compensate for the disadvantage of Otsu algorithm which loses the cerebral vascular details, and obtain the better segmentation and reconstruction results.

Key words: between-class variance, magnetic resonance imaging, cerebrovascular, standard deviation, segmentation, reconstruction

摘要： 磁共振血管造影（magnetic resonance angiography，MRA）是目前临床医学上用以血管成像的首要选择。血管分割的精确程度，是血管三维可视化以及疾病诊断、治疗和评估的关键。对脑血管解剖结构和MRA影像的特性进行了研究，提出了一种基于局部Otsu的脑血管分割方法。首先对原图像进行各向异性扩散滤波；然后通过计算不同的图像分块大小的标准差变化率来确定最佳的分块大小，将图像划分为若干个子图像；用Otsu算法分割子图像，并合并为最终的分割结果；最后利用可视化工具包（visualization toolkit，VTK）重建分割结果，并投影为二维图像进行量化评价。实验结果表明，该方法能够弥补Otsu算法分割MRA脑血管时末端细节丢失的不足，得到更好的分割和重建效果。

关键词: 类间方差, 磁共振造影, 脑血管, 标准差, 分割, 重建