基于椭圆形状约束的Snake模型分割X线图像椎弓根影  

Segmenting Pedicles in Radiographs Based on Snake Model under Elliptical Shape Constraint

作　　者：张俊华[1] LOU Edmond 汪源源[3] 

机构地区：[1]云南大学电子工程系,昆明650091 [2]Rehabilitation Technology Department,Glenrose Rehabilitation Hospital [3]复旦大学电子工程系,上海200433

出　　处：《中国生物医学工程学报》2009年第4期496-500,共5页Chinese Journal of Biomedical Engineering

基　　金：云南省应用基础研究面上项目(2008CD079)

摘　　要：提出一种改进的梯度矢量流(GVF)Snake模型算法用于分割脊椎X线图像中的椎弓根影,并根据分割结果测量脊椎旋转度,以减少手工测量的误差。根据椎弓根影近似椭圆形状的特点,将采用距离函数表示的椭圆形状先验信息作为约束力对Snake的形变过程进行约束,其中的约束椭圆通过对初始分割结果进行拟合获得。迭代进行Snake形变过程和椭圆拟合过程,直到分割结果收敛。基于分割所得椎弓根影边界,计算其中心,进而完成脊椎旋转角度的测量。对208个脊椎X线图像,采用本方法均可准确实现原GVF Snake模型难以完成的对椎弓根影的分割。对手工测量和计算机测量结果分别进行观察者间的误差分析,结果表明:计算机测量减少了观察者间误差,本方法将有助于医生更可靠地测量脊椎旋转度。A modified gradient vector flow （GVF） snake model was proposed to segment pedicles in vertebral radiographs. From segmented pedicles, the vertebral rotation angle was measured to reduce the variability in the manual measurement. Since pedicles were oval-shaped, the elliptical shape prior expressed as the distance between the snake curve and the ellipse was used to constrain the evolution of the snake. The ellipse was adaptively estimated from the snake curve. The snake evolution and the ellipse estimation were iterated until the convergence. Based on detected pedicle boundaries, pedicle centers were identified and the vertebral rotation angle was then calculated. Tested on 208 vertebral radiographs, the proposed method successfully segmented pedicles while the traditional GVF Snake failed to do. The inter-observer variability was analyzed respectively for the manual and computer measurements. It was showed that the inter-observer variability was reduced using the computer method. Therefore, the proposed method may help surgeons measure the vertebral rotation more reliably.

关 键 词：梯度矢量流Snake模型 椭圆形状先验 椎弓根 椎体旋转 X线图像 

分 类 号：R318[医药卫生—生物医学工程]