Schröder Simon, Leydolt Christina, Menapace Rupert, Eppig Timo, Langenbucher Achim
Institute of Experimental Ophthalmology, Saarland University, Homburg, Germany.
University Eye Clinic, Vienna, Austria.
PLoS One. 2016 Jul 8;11(7):e0158988. doi: 10.1371/journal.pone.0158988. eCollection 2016.
The capabilities of a weighted least squares approach for the optimization of the intraocular lens (IOL) constants for the Haigis formula are studied in comparison to an ordinary least squares approach. The weights are set to the inverse variances of the effective optical anterior chamber depth. The effect of random measurement noise is simulated 100000 times using data from N = 69 cataract patients and the measurement uncertainty of two different biometers. A second, independent data set (N = 33) is used to show the differences that can be expected between both methods. The weighted least squares formalism reduces the effect of measurement error on the final constants. In more than 64% it will result in a better approximation, if the measurement errors are estimated correctly. The IOL constants can be calculated with higher precision using the weighted least squares method.
与普通最小二乘法相比，研究了加权最小二乘法优化用于Haigis公式的人工晶状体（IOL）常数的能力。权重设置为有效光学前房深度的逆方差。使用来自N = 69例白内障患者的数据和两种不同生物测量仪的测量不确定性，对随机测量噪声的影响进行了100000次模拟。第二个独立数据集（N = 33）用于展示两种方法之间可能预期的差异。加权最小二乘法形式减少了测量误差对最终常数的影响。如果测量误差估计正确，在超过64%的情况下，它将产生更好的近似值。使用加权最小二乘法可以更精确地计算IOL常数。
