摘要在颅脑血管相关疾病的预防和诊断中，基于放射影像的血管分割是实现其准确诊断的重要前提。从前基于二维影像进行人工勾画的血管轮廓被认为费时费力且准确率较低，目前基于三维放射影像的半自动或自动血管分割算法在一定程度上提高了分割精度。本文聚焦颅脑血管成像领域较为先进的基于传统算法和与深度学习算法相结合的混合算法两大类图像处理方法，其中基于传统算法的图像处理较为基础且常见，而基于深度学习算法的图像处理则还处在逐步走向成熟阶段。笔者通过列举近几年具有代表性的基于放射影像的颅脑血管分割与提取算法，描述采用的图像信息和算法的准确性、鲁棒性和效率等，在深入了解其现状的基础上展望颅脑血管分割领域未来的发展方向及研究重点。

abstractsIn the prevention and diagnosis of cerebrovascular-related diseases, vascular segmentation based on radiography is an important prerequisite for achieving accurate diagnosis. Former, artificial outlines based on two-dimensional images were considered time-consuming and laborious with low accuracy. Currently, semi-automatic or automatic vessel segmentation algorithms based on three-dimensional radiography have improved segmentation accuracy. This article focuses on two advanced algorithms of image processing methods in the field of brain vessel imaging: traditional algorithms and hybrid algorithms combined with deep learning. Between these algorithms, image processing based on traditional algorithms is more basic and common, whereas image processing based on deep learning is still gradually maturing. By enumerating the representative cerebrovascular segmentation and extraction algorithms based on radiography reported in recent years, the authors describe the utilized image information, accuracy, robustness, and efficiency of each algorithm. Moreover, the authors forecast the future development direction and research focus in the field of cerebrovascular segmentation on the basis of the in-depth understanding of the current situation.