Scholarlink

Background <p>To evaluate the impact of age-related ocular anatomical changes on the accuracy of postoperative vault prediction following implantable collamer lens (ICL) surgery and to identify key predictors for vault change (ΔVault).</p> Methods <p>This retrospective study analyzed 2107 eyes that underwent ICL implantation (model V4c). Machine learning algorithms, including Light Gradient Boosting Machine, XGBoost, Support Vector Regression, Random Forest, and Gradient Boosting Decision Tree, were used to construct two prediction models: a full-age cohort model (18–50 years, <i>n</i> = 2107) and a young cohort model (&lt; 40 years, <i>n</i> = 1788). Furthermore, 426 eyes were stratified into three age-based groups for anatomical and dynamic vault analysis: Young group (18–30 years, <i>n</i> = 188), Middle-aged group (31–39 years, <i>n</i> = 134), and Older group (≥ 40 years, <i>n</i> = 104). The preoperative parameters were measured using ultrasound biomicroscopy. Model performance was evaluated using the root mean square error (RMSE), coefficient of determination (R²), mean absolute error, and mean square error. Anatomical parameters and dynamic postoperative vault changes were compared across the age groups.</p> Results <p>The prediction accuracy of the young cohort model (&lt; 40 years) was significantly superior to that of the full-age cohort model (RMSE: 53.59&#xa0;μm vs. 64.36&#xa0;μm; R²: 0.896 vs. 0.881; <i>P</i> &lt; 0.01). Anatomical analysis revealed a significant stepwise increase in lens thickness (LT) from the Young to the Older group (3.5131 ± 0.1962&#xa0;mm, 3.5860 ± 0.2784&#xa0;mm, 3.9212 ± 0.2470&#xa0;mm, <i>P</i> &lt; 0.001), whereas anterior chamber depth (ACD) progressively decreased (3.2662 ± 0.2877&#xa0;mm, 3.1067 ± 0.1958&#xa0;mm, 3.0440 ± 0.2317&#xa0;mm, <i>P</i> &lt; 0.001). The variability of ΔVault, defined as the difference between vault at 6 months postoperatively and vault at 1&#xa0;day postoperatively, was greatest in the Older group (standard deviation = 122.55&#xa0;μm), intermediate in the Middle-aged group (111.61&#xa0;μm), and smallest in the Young group (67.62&#xa0;μm). Feature importance analysis for ΔVault identified ciliary process length, maximum ciliary body thickness, and LT as the top three predictors.</p> Conclusion <p>The inclusion of patients aged 40 years and older significantly reduces the performance of a universal vault prediction model. Older patients exhibit distinct ocular anatomical characteristics, including increased LT, altered ciliary body parameters, and poorer postoperative vault stability. These findings suggest the need to develop age-specific prediction models that incorporate dynamic anatomical factors to optimize surgical planning.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A comparative study on the influence of age-specific ocular anatomical differences on the prediction of post-ICL surgery vault

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月27日

查看全文
  • Jiayu Pan,
  • Yue Wang,
  • Jifan Wang,
  • Ying Li,
  • Yingshuai Li,
  • Xuejun Fang

摘要

Background

To evaluate the impact of age-related ocular anatomical changes on the accuracy of postoperative vault prediction following implantable collamer lens (ICL) surgery and to identify key predictors for vault change (ΔVault).

Methods

This retrospective study analyzed 2107 eyes that underwent ICL implantation (model V4c). Machine learning algorithms, including Light Gradient Boosting Machine, XGBoost, Support Vector Regression, Random Forest, and Gradient Boosting Decision Tree, were used to construct two prediction models: a full-age cohort model (18–50 years, n = 2107) and a young cohort model (< 40 years, n = 1788). Furthermore, 426 eyes were stratified into three age-based groups for anatomical and dynamic vault analysis: Young group (18–30 years, n = 188), Middle-aged group (31–39 years, n = 134), and Older group (≥ 40 years, n = 104). The preoperative parameters were measured using ultrasound biomicroscopy. Model performance was evaluated using the root mean square error (RMSE), coefficient of determination (R²), mean absolute error, and mean square error. Anatomical parameters and dynamic postoperative vault changes were compared across the age groups.

Results

The prediction accuracy of the young cohort model (< 40 years) was significantly superior to that of the full-age cohort model (RMSE: 53.59 μm vs. 64.36 μm; R²: 0.896 vs. 0.881; P < 0.01). Anatomical analysis revealed a significant stepwise increase in lens thickness (LT) from the Young to the Older group (3.5131 ± 0.1962 mm, 3.5860 ± 0.2784 mm, 3.9212 ± 0.2470 mm, P < 0.001), whereas anterior chamber depth (ACD) progressively decreased (3.2662 ± 0.2877 mm, 3.1067 ± 0.1958 mm, 3.0440 ± 0.2317 mm, P < 0.001). The variability of ΔVault, defined as the difference between vault at 6 months postoperatively and vault at 1 day postoperatively, was greatest in the Older group (standard deviation = 122.55 μm), intermediate in the Middle-aged group (111.61 μm), and smallest in the Young group (67.62 μm). Feature importance analysis for ΔVault identified ciliary process length, maximum ciliary body thickness, and LT as the top three predictors.

Conclusion

The inclusion of patients aged 40 years and older significantly reduces the performance of a universal vault prediction model. Older patients exhibit distinct ocular anatomical characteristics, including increased LT, altered ciliary body parameters, and poorer postoperative vault stability. These findings suggest the need to develop age-specific prediction models that incorporate dynamic anatomical factors to optimize surgical planning.