一种基于眼部生物特征参数的外周散焦眼镜个性化拟合预测模型

摘要(中文译文)

构建基于眼部生物测量参数的周边离焦眼镜设计个性化推荐预测模型。本回顾性队列研究纳入2824名在北京大学人民医院眼科中心佩戴周边离焦眼镜的近视儿童。根据眼镜设计将参与者分为五组（多样化分段离焦优化[DSDO]、扩散光学技术[DOT]、离焦诱导多段[DIMS]、高度非球面微透镜[HALT]、六边形优化网状整合[HORI]），并按性别、年龄和基线生物测量参数进行匹配。数据集随机分为训练集（80%）和验证集（20%）用于模型评估。针对每种设计分别构建基于眼部生物测量参数的多元线性回归预测模型。最后，开发了一个R语言函数用于临床个性化推荐。基线参数匹配后，五种周边离焦眼镜设计在眼轴延长控制效果方面未观察到统计学显著差异（P = 0.28）。决定系数（R）在训练集中为0.89至0.96，在验证集中为0.88至0.97；均方根误差（RMSE）在训练集中为0.05 mm至0.07 mm，在验证集中为0.04 mm至0.07 mm。校准图显示预测值与观测值之间具有良好一致性。尽管五种设计的平均效果未观察到显著差异，但本研究成功建立了基于眼部生物测量参数的预测模型，以指导周边离焦眼镜的个性化验配。

Abstract

To construct a predictive model for personalized recommendation of peripheral defocus spectacle designs based on ocular biometric parameters. This retrospective cohort study included 2824 myopic children who wore peripheral defocus spectacles at the Ophthalmology Center of Peking University People's Hospital. Participants were divided into five groups according to the spectacle designs (Diversified Segmental Defocus Optimization [DSDO], Diffusion Optics Technology [DOT], Defocus Induced Multiple Segment [DIMS], Highly Aspherical Lenslets [HALT], Hexagon Optimized Reticular Integration [HORI]) and matched for gender, age, and baseline biometric parameters. The dataset was randomly split into training (80%) and validation (20%) sets for model evaluation. Separate multiple linear regression prediction models based on ocular biometric parameters were constructed for each designs. Finally, an R language function was developed for clinical personalized recommendation. After matching baseline parameters, no statistically significant difference was observed in axial elongation control efficacy among the five peripheral defocus spectacles designs (P = 0.28). The coefficient of determination (R) ranged from 0.89 to 0.96 in the training set and from 0.88 to 0.97 in the validation set; the root mean square error (RMSE) ranged from 0.05 mm to 0.07 mm in the training set and from 0.04 mm to 0.07 mm in the validation set. Calibration plots showed good agreement between predicted and observed values. Although no significant differences in average efficacy were observed among the five designs, this study successfully established predictive models based on ocular biometric parameters to guide personalized fitting of peripheral defocus spectacles.

如何引用