Console J W, Sakata L M, Aung T, Friedman D S, He M
Zhongshan Ophthalmic Center, Guangzhou 510060, People's Republic of China.
Br J Ophthalmol. 2008 Dec;92(12):1612-6. doi: 10.1136/bjo.2007.129932. Epub 2008 Jul 10.
To describe a new image analysis method and software for anterior chamber images obtained by the anterior-segment optical coherence tomography (AS-OCT) and to assess its intraobserver and interobserver measurement reproducibility.
Twenty 8-bit greyscale 600x300 AS-OCT images with apparent wide angles and 20 images with apparent narrow angles were consecutively selected from a database. Two glaucoma fellowship-trained ophthalmologists used proprietary image analysis software to analyse the images twice. Algorithms defined the borders and curvatures of anterior chamber (AC) structures and measured AC parameters using scleral spur location as the only observer input. The intraobserver and interobserver reproducibility of scleral spur location and angle parameters was calculated in terms of limits of agreement (LOA; mean of differences+/-1.96 SD of differences) and coefficient of variation (CV; SD of differences/overall mean).
The analysis software successfully measured all parameters in all images. When the same image was assessed twice by the same grader, the mean differences ranged from 0 to 0.010 mm in linear measurements and 0.001 to 0.006 mm2 in angle area measurements. LOA tended to be greater in the wider angles. The upper and lower limit values of LOAs were approximately 1/5 to 1/4 of the overall mean. Measurements between two graders had a higher variance. Reproducibility in terms of CV was better in wide angles when compared with narrow angles. The reproducibility of scleral spur placement between observers was poorer in narrow angles (p = 0.001). About 50% of the interobserver variance in angle-area measurements was attributable to the variance of scleral spur placement, while this proportion was only 10-20% in linear measurements.
Determination of angle parameters using semiautomated software leads to variability in measurement. Variability increases when more than one observer identifies the scleral spur. Variability differs in narrow and open angles, and so including both types is essential when evaluating angle-assessment software. A fully automated analysis and higher image resolution would likely improve quantification of Visante AS-OCT images.
描述一种用于分析前段光学相干断层扫描（AS-OCT）获得的前房图像的新图像分析方法和软件，并评估其在观察者内和观察者间测量的可重复性。
从数据库中连续选取20张8位灰度、600x300且前房角明显较宽的AS-OCT图像以及20张前房角明显较窄的图像。两位接受过青光眼专科培训的眼科医生使用专用图像分析软件对图像进行两次分析。算法通过巩膜突位置作为唯一观察者输入来定义前房（AC）结构的边界和曲率，并测量AC参数。观察者内和观察者间巩膜突位置及角度参数的可重复性通过一致性界限（LOA；差异均值±1.96倍差异标准差）和变异系数（CV；差异标准差/总体均值）来计算。
分析软件成功测量了所有图像中的所有参数。当同一位分级者对同一图像进行两次评估时，线性测量的平均差异范围为0至0.010毫米，角面积测量的平均差异范围为0.001至0.006平方毫米。在较宽的前房角中，LOA往往更大。LOA的上限和下限值约为总体均值的1/5至1/4。两位分级者之间的测量差异更大。与窄角相比，宽角的CV可重复性更好。观察者之间巩膜突位置在窄角中的可重复性较差（p = 0.001）。角面积测量中观察者间差异的约50%归因于巩膜突位置的差异，而在线性测量中这一比例仅为10 - 20%。
使用半自动软件确定角度参数会导致测量变异性。当有不止一位观察者识别巩膜突时，变异性会增加。窄角和开角中的变异性不同，因此在评估角度评估软件时纳入这两种类型至关重要。全自动分析和更高的图像分辨率可能会改善Visante AS-OCT图像的量化。
