基于对称区域生长和边缘梯度的视神经纤维的分割

Segmentation of Optic Nerve Fiber Based on Symmetric Region Growing and Edge Gradient

在视神经横切面图像中,将每个神经纤维的内外边界进行精确分割是视神经形态分析的重要环节,提出一种基于对称区域生长和髓鞘边缘梯度的有效分割算法。该算法分两步进行,首先根据交互方式下选取的种子点,由对称区域生长算法实现轴突分割,然后在轴突轮廓模型基础上,髓鞘外轮廓在髓鞘平均边缘梯度引导下进行演化,实现自动分割。与K-均值聚类,局部阈值和水平集等其他算法的实验结果相对照显示,该算法分割获得的轴突和髓鞘轮廓与实际轮廓相吻合,其分割结果可以作为后续神经纤维形态分析的基础。

The segmentation of nerve fiber in the transverse section of optic nerve is an important procedure for optic nerve morphometry. We proposed an efficient approach for the segmentation based on symmetric region growing and average edge gradient. Symmetric region growing algorithm was applied to segment axons according to the seed points which were chosen interactively. Then, based on the initial contour provided by the axon model, the outer contour of myelin sheath leaded by the average edge gradient was detected automatically. Experimental result demonstrated that the axon and myelin sheath boundaries matched the real one well. The proposed method was efficient compared to the other methods including K-means clustering, local threshold and levelset.