Mack Lauren M, Kim Sung Yoon, Lee Sungmin, Sangi-Haghpeykar Haleh, Lee Wesley
Division of Women's and Fetal Imaging, Department of Obstetrics and Gynecology, Baylor College of Medicine, Houston, Texas USA
Samsung Medison Research and Development Center, Seoul, Korea.
J Ultrasound Med. 2016 Jul;35(7):1573-8. doi: 10.7863/ultra.15.09086. Epub 2016 Jun 7.
The purpose of this study was to document the reproducibility and efficiency of a semiautomated image analysis tool that rapidly provides fetal fractional limb volume measurements. Fifty pregnant women underwent 3-dimensional sonographic examinations for fractional arm and thigh volumes at a mean menstrual age of 31.3 weeks. Manual and semiautomated fractional limb volume measurements were calculated, with the semiautomated measurements calculated by novel software (5D Limb Vol; Samsung Medison, Seoul, Korea). The software applies an image transformation method based on the major axis length, minor axis length, and limb center coordinates. A transformed image is used to perform a global optimization technique for determination of an optimal limb soft tissue boundary. Bland-Altman analysis defined bias with 95% limits of agreement (LOA) between methods, and timing differences between manual versus automated methods were compared by a paired t test. Bland-Altman analysis indicated an acceptable bias with 95% LOA between the manual and semiautomated methods: mean arm volume ± SD, 1.7% ± 4.6% (95% LOA, -7.3% to 10.7%); and mean thigh volume, 0.0% ± 3.8% (95% LOA, -7.5% to 7.5%). The computer-assisted software completed measurements about 5 times faster compared to manual tracings. In conclusion, semiautomated fractional limb volume measurements are significantly faster to calculate when compared to a manual procedure. These results are reproducible and are likely to reduce operator dependency. The addition of computer-assisted fractional limb volume to standard biometry may improve the precision of estimated fetal weight by adding a soft tissue component to the weight estimation process.
本研究的目的是记录一种能快速提供胎儿肢体分数体积测量值的半自动图像分析工具的可重复性和效率。50名孕妇在平均月经年龄31.3周时接受了三维超声检查,以测量手臂和大腿的分数体积。计算了手动和半自动的肢体分数体积测量值,半自动测量值由新型软件(5D Limb Vol;韩国首尔三星麦迪逊公司)计算得出。该软件应用基于长轴长度、短轴长度和肢体中心坐标的图像变换方法。通过变换后的图像执行全局优化技术,以确定最佳肢体软组织边界。采用Bland-Altman分析定义两种方法之间的偏差及95%一致性界限(LOA),并通过配对t检验比较手动与自动方法之间的时间差异。Bland-Altman分析表明,手动和半自动方法之间的偏差及95% LOA可接受:平均手臂体积±标准差,1.7%±4.6%(95% LOA,-7.3%至10.7%);平均大腿体积,0.0%±3.8%(95% LOA,-7.5%至7.5%)。与手动描记相比,计算机辅助软件完成测量的速度快约5倍。总之,与手动操作相比,半自动肢体分数体积测量的计算速度明显更快。这些结果具有可重复性,且可能减少对操作人员的依赖。在标准生物测量中增加计算机辅助的肢体分数体积测量,可能通过在体重估计过程中加入软组织成分来提高胎儿体重估计的精度。