基于3D方向可控制滤波器的MRA脑血管检测

Detection of Cerebral Vessels in MRA Based on 3D Steerable Filters

针对低信噪比的磁共振造影(MRA)图像,本文提出了一种新的基于方向可控制滤波器的3D脑血管增强和检测方法.我们把3D二进B样条小波作为基滤波器,通过线性组合得到一个可旋转到任意角度的3D方向滤波器.基于该滤波器响应的方向能量被用来检测血管的方向强度.对多尺度变换下得到的局部能量最大值图像,采用简单的阈值分割就能抽出完整的脑血管树.我们对实际病人的MRA体数据进行了测试,得到了比较满意的结果.实验结果表明,我们提出的方法简单、可靠,并适合于其他管状结构如心血管、支气管树的检测和抽取.

A fully automatic method for enhancement and segmentation of three-dimensional cerebral vessels in low contrast MRA is presented. We obtain the 3D dyadic B-spline wavelets by extending the corresponding 1D wavelet. A 3D steerable filter is then developed based on 3D dyadic B-spline wavelets,it can be adapted to an arbitrary direction. The oriented energy of filter response is introduced for detecting orientation strength of vessels in that direction. The points with maximum of local oriented energy across multiple scales are detected and then cerebral vessel tree can be extracted by simple thresholding. This method was tested on real MRA data and promising results have been obtained. It could be suitable for other types of curvilinear structures such as cardiovascular vessels,bronchial tree.