七种基于人工智能的人工晶状体度数计算公式在极长眼轴高加索人眼中的准确性
该回顾性准确性分析在乌克兰基辅临床眼科医院眼显微手术中心纳入48只眼轴>30.00 mm的高度近视眼,比较7种基于人工智能的人工晶状体度数计算公式。术后3个月测量屈光度,Hill-RBF 3.0的均方根绝对误差最低(0.788),仅对Karmona(0.956)有统计学优势;Hoffer QST(0.442)和Hill-RBF(0.490)的中位绝对误差显著优于LSF AI(0.800);Hill-RBF 3.0、Kane和Hoffer QST的预测误差在±0.50 D以内的眼占比最高(各54.17%)。
要点
- 48只眼轴>30.00 mm眼,Hill-RBF 3.0均方根绝对误差最低(0.788),仅显著优于Karmona(0.956,P=0.021)。
- 中位绝对误差Hoffer QST 0.442和Hill-RBF 0.490显著优于LSF AI 0.800(P=0.013和P=0.008)。
- 预测误差±0.50 D以内占比Hill-RBF 3.0、Kane、Hoffer QST各54.17%,显著高于LSF AI 27.08%和Karmona 39.58%。
核心结论
- 所有测试公式均表现出可比的准确性,Hill-RBF 3.0在均方根绝对误差上比Karmona更准确,LSF AI在中位绝对误差上不如Hoffer QST和Hill-RBF 3.0准确。
研究设计与样本
- 回顾性准确性和有效性分析,在乌克兰基辅临床眼科医院眼显微手术中心开展,纳入48只眼轴>30.00 mm的高度近视眼,均行无缝超声乳化术。
主要发现
- Hill-RBF 3.0的均方根绝对误差最低(0.788),仅对Karmona(0.956)有统计学优势(P=0.021)。
- 中位绝对误差Hoffer QST为0.442、Hill-RBF为0.490,分别显著优于LSF AI的0.800(P=0.013和P=0.008)。
- 预测误差在±0.50 D以内的眼占比Hill-RBF 3.0、Kane和Hoffer QST均为54.17%,显著高于LSF AI的27.08%和Karmona的39.58%;Hoffer QST显著高于LSF AI。
关键数值与推荐
- 主要结局指标为均方根绝对误差、中位绝对误差和预测误差在±0.50 D以内的眼百分比。
- 术前人工晶状体度数从SRK/T、Holladay 2或Barrett Universal II的结果中随机选择。
- 术后人工晶状体度数计算使用Hill-RBF 3.0、Kane、PEARL-DGS、Ladas Super Formula AI(LSF AI)、Hoffer QST、Karmona和Zhu-Lu公式。
临床怎么用
- 原文未据此给出推荐。
局限性
- 原文未报告局限性。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
PURPOSE: To compare 7 artificial intelligence (AI)-based intraocular lens (IOL) power calculation formulas in extremely long eyes. DESIGN: Retrospective accuracy and validity analysis. SETTING: Kyiv Clinical Ophthalmology Hospital Eye Microsurgery Center, Ukraine. STUDY POPULATION: Patients with highly myopic eyes, who underwent uneventful phacoemulsification. OBSERVATION PROCEDURES: IOL power was calculated before cataract surgery. The power of the implanted IOL was randomly selected from the outcomes of SRK/T, Holladay 2, or Barrett Universal II. Three months after phacoemulsification, refraction was measured. Postsurgery IOL power calculations were performed using the following formulas: Hill-RBF 3.0, Kane, PEARL-DGS, Ladas Super Formula AI (LSF AI), Hoffer QST, Karmona, and Zhu-Lu. MAIN OUTCOME MEASURES: Root mean square absolute error (RMSAE), median absolute error (MedAE), and percentage of eyes with prediction error within ±0.50 D. RESULTS: Forty-eight eyes with axial length >30.00 mm were studied. Hill-RBF 3.0 yielded the lowest RMSAE (0.788) with statistical superiority only over Karmona (0.956, P = .021). In terms of MedAE, outcomes obtained by Hoffer QST (0.442) and Hill-RBF (0.490) were statistically significant compared with LSF AI (0.800, P = .013 and P = .008, respectively). The highest percentage of eyes with prediction error within ±0.50 D was achieved by Hill-RBF 3.0, Kane, and Hoffer QST (54.17% each) statistically significant as follows: both Hill-RBF and Kane compared with LSF AI (27.08%) and Karmona (39.58%), and Hoffer QST compared with LSF AI. CONCLUSION: All tested formulas demonstrated comparable trueness, with Hill-RBF 3.0 being more accurate than Karmona (RMSAE), and LSF AI being less accurate than Hoffer QST and Hill-RBF 3.0 (MedAE).
引用
- Stopyra W, Voytsekhivskyy O, Grzybowski A, et al. Accuracy of 7 Artificial Intelligence-Based Intraocular Lens Power Calculation Formulas in Extremely Long Caucasian Eyes. Am J Ophthalmol. 2024. PMID: 39536848. doi:10.1016/j.ajo.2024.10.033
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。