Department of Electronic Engineering, Yunnan University, Kunming, 650091, China.
Comput Med Imaging Graph. 2009 Dec;33(8):602-7. doi: 10.1016/j.compmedimag.2009.06.002. Epub 2009 Jul 10.
A modified graph cut was proposed under the elliptical shape constraint to segment cervical lymph nodes on sonograms, and its effect on the measurement of short axis to long axis ratio (S/L) was investigated by using the relative ultimate measurement accuracy (RUMA). Under the same user inputs, the proposed algorithm successfully segmented all 60 sonograms tested, while the traditional graph cut failed. The mean RUMA resulted from the developed method was comparable to that resulted from the manual segmentation. Results indicated that utilizing the elliptical shape prior could appreciably improve the graph cut for nodes segmentation, and the proposed method satisfied the accuracy requirement of S/L measurement.
提出了一种在椭圆形状约束下的改进图割算法,用于对超声图像中的颈淋巴结进行分割,并使用相对极限测量精度(RUMA)来研究其对短轴与长轴比(S/L)测量的影响。在相同的用户输入下,所提出的算法成功地分割了所有 60 个测试的超声图像,而传统的图割算法则失败了。该方法的平均 RUMA 与手动分割的结果相当。结果表明,利用椭圆形状先验可以显著提高图割算法对节点的分割效果,所提出的方法满足 S/L 测量的精度要求。