基于最佳方向性梯度通量血管增强的冠脉分割

Coronary Arteries Segmentation Based on Optimal Oriented Flux Vessel Enhancement

针对多层螺旋CT(MSCT)冠状动脉分割时受周围静脉血管等组织的影响而容易发生泄漏的问题,提出了一种基于最佳方向性梯度通量(OOF)血管增强的分割方法.首先,得到原始图像的梯度向量场,选择合适的半径,计算球面特定方向上投影梯度的通量,寻找使得流向球体内部的投影通量最小的最佳方向.求解最佳方向上梯度通量矩阵的特征值,利用特征值构造血管相似度响应函数,对冠状动脉进行增强,之后采用自适应阈值的区域生长方法将冠脉血管分割出来.实验结果表明,该算法受冠脉周围组织的影响较小,避免了泄漏的发生,而且能提取到较多的细小分支.

A segmentation algorithm based on optimal oriented flux （OOF） vessel enhancement was pro- posed to extract coronary arteries in cardiac multislice computed tomography （MSCT） images by elimina- ting influence of nearby structures such as vein. First, the gradient vector field was calculated, and then a series of radiuses were chosen to compute the OOF on the sphere surface. The optimal direction in which the inward projected flux is minimal coincides with that in which the vessel lies in, so the eigenval- ues can be calculated with which vessel similarity function will be constructed to enhance the original da- ta. Hereafter, an adaptive threshold region growing is executed to extract the coronary arteries. Experi- mental results illustrate that the algorithm is not easily influenced by the structure around the arteries and thus over-segmentation can be avoided. Simultaneously, more thin branches can be obtained.