Wei Ling, Song Yunxiao, He Wenwen, Chen Xu, Ma Bo, Lu Yi, Zhu Xiangjia
Department of Ophthalmology and Eye Institute, Eye & ENT Hospital, Fudan University, Shanghai, China.
National Health Commission Key Laboratory of Myopia, Fudan University, Shanghai, China.
Front Med (Lausanne). 2020 Dec 23;7:592663. doi: 10.3389/fmed.2020.592663. eCollection 2020.
To develop a machine learning-based calculator to improve the accuracy of IOL power predictions for highly myopic eyes. Data of 1,450 highly myopic eyes from 1,450 patients who had cataract surgeries at our hospital were used as internal dataset (train and validate). Another 114 highly myopic eyes from other hospitals were used as external test dataset. A new calculator was developed using XGBoost regression model based on features including demographics, biometrics, IOL powers, A constants, and the predicted refractions by Barrett Universal II (BUII) formula. The accuracies were compared between our calculator and BUII formula, and axial length (AL) subgroup analysis (26.0-28.0, 28.0-30.0, or ≥30.0 mm) was further conducted. The median absolute errors (MedAEs) and median squared errors (MedSEs) were lower with the XGBoost calculator (internal: 0.25 D and 0.06 D; external: 0.29 D and 0.09 D) vs. the BUII formula (all ≤ 0.001). The mean absolute errors and were 0.33 ± 0.28 vs. 0.45 ± 0.31 (internal), and 0.35 ± 0.24 vs. 0.43 ± 0.29 D (external). The mean squared errors were 0.19 ± 0.32 vs. 0.30 ± 0.36 (internal), and 0.18 ± 0.21 vs. 0.27 ± 0.29 D (external). The percentages of eyes within ±0.25 D of the prediction errors were significantly greater with the XGBoost calculator (internal: 49.66 vs. 29.66%; external: 78.28 vs. 60.34%; both < 0.05). The same trend was in MedAEs and MedSEs in all subgroups (internal) and in AL ≥30.0 mm subgroup (external) (all < 0.001). The new XGBoost calculator showed promising accuracy for highly or extremely myopic eyes.
开发一种基于机器学习的计算器，以提高高度近视患者人工晶状体（IOL）屈光度预测的准确性。将我院1450例接受白内障手术的高度近视患者的1450只眼睛的数据用作内部数据集（训练和验证）。另外114只来自其他医院的高度近视眼睛用作外部测试数据集。基于人口统计学、生物特征、IOL屈光度、A常数以及Barrett通用II（BUII）公式预测的屈光不正等特征，使用XGBoost回归模型开发了一种新的计算器。比较了我们的计算器和BUII公式之间的准确性，并进一步进行了眼轴长度（AL）亚组分析（26.0 - 28.0、28.0 - 30.0或≥30.0 mm）。与BUII公式相比，XGBoost计算器的中位绝对误差（MedAEs）和中位平方误差（MedSEs）更低（内部：0.25 D和0.06 D；外部：0.29 D和0.09 D），而BUII公式的相应误差均≤0.001）。平均绝对误差分别为0.33±0.28与0.45±0.31（内部），以及0.35±0.24与0.43±0.29 D（外部）。平均平方误差分别为0.19±0.32与0.30±0.36（内部），以及0.18±0.21与0.27±0.29 D（外部）。XGBoost计算器预测误差在±0.25 D范围内的眼睛百分比显著更高（内部：49.66对29.66%；外部：78.28对60.34%；均P < 0.05）。所有亚组（内部）以及AL≥30.0 mm亚组（外部）的MedAEs和MedSEs均呈现相同趋势（均P < 0.001）。新的XGBoost计算器在高度或超高度近视眼中显示出有前景的准确性。
