摘要
目的基于眼生物参数序列建立一种屈光度预测方法。方法采用分层随机整群抽样方法抽取2022年1月至2023年1月复旦大学附属眼耳鼻喉科医院采集的上海杨浦区2所重点学校和2所普通学校5~13岁儿童的数据集。收集儿童的眼生物参数数据包括性别、年龄、屈光度、眼轴长、角膜曲率和前房深度信息,采用最小二乘法计算最优拟合直线的斜率,结合基线数据及预处理后的眼生物参数变化率,建立最小二乘-反向传播(BP)神经网络模型。研究中将数据集依照8∶2的比例划分为训练集和验证集进行五折交叉验证,并采用平均绝对误差(MAE)、均方误差(MSE)、均方根误差(RMSE)、相关系数R和决定系数R 2指标评估模型性能。结果最小二乘-BP神经网络模型的性能指标R^(2)、R、RMSE、MAE、MSE分别为0.96、0.9819、0.2142、0.1399 D、0.0459,均达到最佳值。屈光度预测值与真实值之间的回归方程为y=0.97 x+0.0148,R^(2)=0.97,具有良好的相关性。在内部验证中,随访3、6、9、12个月的MAE分别为0.1101、0.1360、0.1537、0.1848 D,均达到了临床可接受性能(小于0.25 D)。在外部验证中,所有年龄段的误差均小于0.25 D。结论建立了一种基于眼生物参数序列的屈光度预测方法。
Objective To establish a prediction method of diopter based on sequence of ocular biological parameters.Methods A stratified random cluster sampling method was used to extract the dataset.The dataset consisted of data collected from January 2022 to January 2023 by the Eye&ENT Hospital,Fudan University,from children aged 5 to 13 years in 2 key schools and 2 general schools of Yangpu District,Shanghai.Children’s ocular biological parameters,including sex,age,diopter,axial length,corneal curvature,and anterior chamber depth were collected.The slope of the optimally fitted straight line was calculated using the least squares method.The least square-back propagation(BP)neural network model was established by combining baseline data and the pre-processed rate of the change of ocular biological parameters.The dataset was divided into the training set and the validation set according to the ratio of 8:2 for five-fold cross-validation.The model performance was evaluated by using the mean absolute error(MAE),mean squared error(MSE),root mean square error(RMSE),correlation coefficient R,and coefficient of determination R^(2).Results The optimal performances of R^(2),R,RMSE,MAE,and MSE of the least square-BP neural network model were 0.96,0.9819,0.2142,0.1399 D,0.0459,respectively.The regression equation between the predicted value and the true value of the diopter was y=0.97 x+0.0148,R^(2)=0.97,with good correlation.In the internal verification,MAE values of the diopter at three,six,nine,and twelve months of follow-up were 0.1101,0.1360,0.1537,and 0.1848 D,respectively,which achieved clinically acceptable performance(less than 0.25 D).In the external validation,the errors were less than 0.25 D at all ages.Conclusions A prediction method of diopter based on sequence of ocular biological parameters was successfully developed.
作者
徐略标
丁岚
梁宸
王钰靓
刘雨佳
尚建慜
朱俊
项华中
褚仁远
王成
瞿小妹
Xu Luebiao;Ding Lan;Liang Chen;Wang Yuliang;Liu Yujia;Shang Jianmin;Zhu Jun;Xiang Huazhong;Chu Renyuan;Wang Cheng;Qu Xiaome(Institute of Biomedical Optics and Optometry,Key Lab of Medical Optical Technology and Instruments,Ministry of Education,School of Health Science and Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China;Shanghai Optometry Research Center,Key Laboratory of Myopia,Chinese Academy of Medical Sciences and National Health Commission,Department of Ophthalmology,Eye&ENT Hospital,Fudan University,Shanghai 200030,China)
出处
《国际生物医学工程杂志》
CAS
2024年第5期417-422,共6页
International Journal of Biomedical Engineering
基金
国家自然科学基金(82171093)
上海市科研计划(21dz2311300)
闵行区自然科学基金(00222000051)。
关键词
近视
眼生物参数
屈光度
最小二乘法
反向传播神经网络
Myopia
Ocular biological parameters
Diopter
Least square method
Back propagation neural network
