三维窄带水平集算法在内耳分割中的应用

Three-dimensional narrow-band level set algorithm for inner ear segmentation

为减少从颞骨螺旋CT(computed tomography)图像中分割内耳的手动交互量,阐述了基于区域竞争的窄带水平集算法的基本原理及其特点,构造了合适的速度函数来控制水平集函数的演化,并将该算法应用于颞骨螺旋CT图像中内耳的分割.通过建立三维模型表面上的点与3个正交截面的对应关系,可以快速确定弱边界所在的位置,然后在正交截面上手动编辑以去除多余的组织,得到完整的内耳.对3个病人的5例颞骨图像进行内耳分割实验,实现了4例内耳的完整分割、1例严重畸形内耳的不完整分割.分割1例内耳约需10 min.该方法具有分割速度快、手动交互少、结果表面均匀的优点.

3D narrow-band level set algorithm based on region competition was modified and applied for the segmentation of inner ear to minimize the manual operation for segmenting inner ear from spiral computed tomography （CT） images of temporal bones. The basic principle and characteristics of narrow-band level set algorithm based on region competition were described. An effective speed function was designed to control the evolution of the level set function. The 3D narrow-band level set algorithm was successfully used to separate the inner ear from spiral CT images of the temporal bone. The correlation of the point on the resultant surface to the three orthogonal sections intersecting at this point was built to quickly find the weak edges. Then the unwanted structures were deleted manually by viewing orthogonal sections slice by slice. The method was tested by segmenting five inner ears of three patients, four of which were separated completely and one whose shape was apparently abnormal contained more than the objects of interest. The segmentation method took about 10 min to obtain a complete inner ear model. Characteristics of the method are fast, simple interaction and the result surface is even.