Smith R Theodore, Nagasaki Takayuki, Sparrow Janet R, Barbazetto Irene, Koniarek Jan P, Bickmann Lee J
Department of Ophthalmology, Columbia University, New York, NY 10032, USA.
J Biomed Opt. 2004 Jan-Feb;9(1):162-72. doi: 10.1117/1.1630604.
Normal macular photographic patterns are geometrically described and mathematically modeled. Forty normal color fundus photographs were digitized. The green channel gray-level data were filtered and contrast enhanced, then analyzed for concentricity, convexity, and radial resolution. The foveal data for five images were fit with elliptic quadratic polynomials in two zones: a central ellipse and a surrounding annulus. The ability of the model to reconstruct the entire foveal data from selected pixel values was tested. The gray-level patterns were nested sets of concentric ellipses. Gray levels increased radially, with retinal vessels changing the patterns to star shaped in the peripheral fovea. The elliptic polynomial model could fit a high-resolution green channel foveal image with mean absolute errors of 6.1% of the gray-level range. Foveal images were reconstructed from small numbers of selected pixel values with mean errors of 7.2%. Digital analysis of normal fundus photographs shows finely resolved concentric elliptical foveal and star-shaped parafoveal patterns, which are consistent with anatomical structures. A two-zone elliptic quadratic polynomial model can approximate foveal data, and can also reconstruct it from small subsets, allowing improved macular image analysis.
对正常黄斑的摄影图像进行了几何描述和数学建模。对40张正常彩色眼底照片进行了数字化处理。对绿色通道的灰度数据进行滤波和对比度增强，然后分析其同心度、凸度和径向分辨率。对五幅图像的中央凹数据在两个区域用椭圆二次多项式进行拟合：一个中央椭圆和一个周围的环带。测试了该模型从选定像素值重建整个中央凹数据的能力。灰度模式是嵌套的同心椭圆集。灰度值沿径向增加，视网膜血管使中央凹周边的模式变为星形。椭圆多项式模型能够拟合高分辨率绿色通道中央凹图像，平均绝对误差为灰度范围的6.1%。从少量选定像素值重建中央凹图像，平均误差为7.2%。对正常眼底照片的数字分析显示，中央凹有精细分辨的同心椭圆形模式，周边有星形模式，这与解剖结构一致。一个双区域椭圆二次多项式模型可以近似中央凹数据，也可以从小子集重建该数据从而改进黄斑图像分析。
