基于多视图信息的CT序列图像颈椎间盘分割算法

Segmentation Algorithm of Cervical Intervertebral Disc in CT Sequence Image based on Multi-view Information

从颈椎CT(电子计算机断层扫描)序列图像中准确地分割出颈椎间盘是颈椎终板三维分型的基础,但在原始水平面数据中,颈椎间盘没有明显的边界,难以实现高精度的颈椎间盘分割。为此,在深入研究颈椎间盘本身的三维空间结构以后,提出了一种结合多视图信息的颈椎间盘分割算法,其能充分利用颈椎间盘上下表面的三维空间信息,解决了颈椎间盘边界在水平面难以界定的问题,使分割得到的颈椎间盘更加精确。从多组颈椎CT序列图像上的实验结果表明,文章提出的算法与双阈值法和大津法相比,实现了精度更高的颈椎间盘分割,更符合实际应用的需求。

Accurate segmentation of cervical intervertebral disc from cervical CT(Computed Tomography) sequence images is the basis of three-dimensional classification of cervical endplate. However, cervical intervertebral disc has no obvious boundary in the original horizontal data, so it is difficult to achieve high precision cervical intervertebral disc segmentation. Therefore, this paper proposes a cervical disc segmentation algorithm based on multi-view information after in-depth study of the three-dimensional spatial structure of the cervical disc itself. This algorithm makes full use of the threedimensional spatial information of the upper and lower surfaces of the cervical disc, solves the problem that the boundary cervical disc is difficult to be defined in the horizontal plane, and makes the segmentation of the cervical intervertebral disc more accurate. The Results of the experiments on multiple sets of cervical vertebra CT sequence images show that the proposed algorithm achieves higher precision segmentation of cervical intervertebral disc, compared with double threshold method and OTSU method. It is more in line with the needs of practical applications.