基于数字图像视觉特征自动检测超声图像中胎儿颈项透明层  

Digital image visual feature-based automatic measurement of nuchal translucency of fetus in ultrasound images

作　　者：丁红[1] 张永[2] 刘颖琳[1] 陈立斌[1] 

机构地区：[1]中山大学孙逸仙纪念医院产科,广州510120 [2]中山大学信息学院

出　　处：《中华生物医学工程杂志》2012年第5期348-352,共5页Chinese Journal of Biomedical Engineering

基　　金：广东省科技计划项目（20098030801177）

摘　　要：目的利用计算机技术设计一种基于数字图像视觉特征的超声图像中胎儿颈项透明层（NT）的自动检测方法。方法建立基于数字图像视觉特征自动检测超声图像中胎儿NT的方法，包括胎儿超声图像提取、连通分量提取、目标检测和NT测量的实现共4个主要步骤。评价标准实验图像（n=35）、临床挑选图像（n=1208）和人工已标注图像（n=120）3组图像采用自动检测方法进行NT测量的实验结果，包括其定位准确性和测量误差，并记录最大误差值。结果应用自动检测方法检测标准实验图像、临床挑选图像和人工已标注图像中胎儿NT的定位准确率分别为100％（35／35）、90．7％（1096／1208）、90．8％（109／120），NT的测量误差均小于0．03mm。结论成功建立了基于数字图像视觉特征自动检测超声图像中胎儿NT的方法，可减少人工测量的主观性和随机性，提高超声筛查的准确性。Objective To design an approach for automatic detection of nuchal translucency（NT） of fetus in ultrasound images based on digital image visual features via computer technology. Methods The established approach for digital image visual feature-based automatic detection of NT of fetus in ultrasound images consisted of image retrieval, connected component extraction, target assessment and measurement of NT. Results of automatic NT measurement were obtained from standard experimental images（n=35）, clinically screened images （n=1208） and artificially marked images （n=120）. The accuracy and bias of positioning were examined and maximal errors were recorded. Results The accuracy of NT positioning was 100% （35/35）, 90.7% （1096/1208） and 90.8% （109/120） corresponding to standard experimental images, clinically screened images and artificially marked images via automatic measurement respectively. An error of less than 0.03 mm was found. Conclusion The approach for automatic measurement of NT of fetus in ultrasound images based on digital image visual feature is successful established, which can reduce the subjectivity and randomness of manual operation and improve accuracy of ultrasound screening.

关 键 词：图像处理 计算机辅助 超声检查 产前 唐氏综合征 颈项透明层 

分 类 号：R714.5[医药卫生—妇产科学]