摘要
为从超声图像中准确估计胎儿股骨长度、头围、双顶径和枕额径,提出一种新的胎儿超声图像分割和径线自动测量方法。首先使用结合空间邻域信息的二维模糊聚类,在抑制斑点噪声的同时提取胎儿超声图像中的骨骼部分;接着将骨骼提取的结果细化作为霍夫变换的输入;然后根据股骨和头部不同的形态特征,分别设计了直线霍夫变换加二次曲线拟合的股骨参数提取方法,以及椭圆二次迭代霍夫变换的头部参数提取方法。本研究对60组胎儿超声图像自动计算所得的股骨长度和头围与手动测量间的差异分别为(0.069±0.066)cm和(-0.508±0.458)cm。结果显示本方法所得结果基本与手动测量的结果吻合,可以提高测量的客观性,有望应用于临床的产前监护。
In order to acquire the accurate measurement of the femur length,head circumference,biparietal diameter and occipitofrontal diameter based on fetal ultrasound images,a novel method was proposed for the fetal ultrasound image segmentation and the measurement.The two dimensional fuzzy clustering method based on the spatial information was firstly used to suppress speckle and extract the part of bone from fetal ultrasound images.Then the extraction result was thinned to serve as the input data for Hough transform.According to the different shape traits of the femur and head,two processing procedure were utilized separately.Hough transform for the line and conic polynomial regression was used to estimate the femur length,and two iterative Hough transform for head parameters.For 60 groups of fetal ultrasound images,the differences between the automatic measurement and manual measurement were(0.069±0.066)cm for the femur length and(-0.508±0.458)cm for the head circumference.The results showed that the proposed method could provide the results consistent with the ones from manual measurement.With the improvement in the objectivity of the measurement,the proposed method has potential in applying to clinical antepartum monitoring.
出处
《中国生物医学工程学报》
CAS
CSCD
北大核心
2007年第6期867-873,883,共8页
Chinese Journal of Biomedical Engineering
基金
国家重点基础研究规划基金(2006CB705707上海市重点学科建设项目(B112)
复旦大学研究生创新基金(EYH1220001)。
关键词
胎儿超声图像分割
二维模糊聚类
随机霍夫变换
曲线拟合
二次迭代霍夫变换
fetal ultrasound image segmentation
two dimensional fuzzy clustering
randomized Hough transform
conic polynomial regression
two iterative Hough transform