基于图搜索和轮廓插值的半自动组织分割方法

Half-automatic Segmentation Algorithm Based on Graph Search and Contour Interpolation

提出了一种基于图搜索和轮廓插值的半自动序列医学图像分割方法,该方法首先采用一种改进的Live-wire方法对医学图像序列中反映待分割组织主要形状结构的关键层进行交互式分割,然后利用轮廓插值技术对中间的过渡图像进行自动分割;对传统的轮廓插值算法进行了改进,采用了基于余弦函数的轮廓插值,在轮廓形状有突变的地方,采用余弦插值可以很好地改变线性插值所带来的弊端,及时改变插值轮廓形状。使得插值生成的边界轮廓和实际边界轮廓的符合度大大提高。最后结合中间层的局部信息通过自动收缩和用户局部调节的办法对中间层分割结果进行修正,从而获得最终的分割结果。实验结果表明,该方法对于复杂的人体组织和器官可以获得满意的分割结果。

Based on graph search and contour interpolation algorithm, a half-automatic algorithm of serial images segmentation is presented. The active control scheme is used to combine some related algorithms. Also traditional contour interpolation and Lave-wire algorithm are modified so that according to the real images local character, the combined active contour model self-shrink for rebuilding the contour. Combining with the ideas of active contour model, the images are self-shrinking on the basis of actual images local character in rebuilt contour.The experiment shows the algorithm can obtain the region of interest from a series of medical images quickly and reliably.