Related Experiment Video
Updated: Jan 17, 2026
09:10
Digital Hybrid Model Preparation for Virtual Planning of Reconstructive Dentoalveolar Surgical Procedures
Published on: August 5, 2021
2.2K
Using Artificial Intelligence to Predict Implantable Collamer Lens Vault: A Low Parameter-Dependent Model for Better
Peien Sheng1, Yinan Liu1, Mingyue Shen1
1Department of Ophthalmology, Peking University Third Hospital, 49th North Garden Road, Haidian District, Beijing, People's Republic of China.
Translational Vision Science & Technology
|September 23, 2025
Summary
Artificial intelligence accurately predicts implantable collamer lens (ICL) vault, identifying key influencing factors. This AI model enhances ICL surgery safety by minimizing abnormal postoperative vaults.
Area of Science:
- Ophthalmology
- Artificial Intelligence
- Biomedical Engineering
Background:
- Accurate prediction of the implantable collamer lens (ICL) vault is crucial for successful refractive surgery.
- Traditional methods often struggle with the complex, nonlinear relationships between preoperative parameters and postoperative vault.
Purpose of the Study:
- To develop and validate an artificial intelligence (AI) model for predicting ICL vault.
- To interpret the contributions of various preoperative parameters to ICL vault using AI.
Main Methods:
- Extreme Gradient Boosting (XGBoost) machine learning algorithm was employed.
- A dataset of 247 eyes was used for training and testing, with an additional 50 eyes for external validation.
- SHapley Additive exPlanations (SHAP) was utilized for model interpretability.
Main Results:
- The AI model demonstrated robust performance in predicting ICL vault, with low error metrics on both test and validation sets.
- Key parameters influencing vault included horizontal sulcus-to-sulcus distance (STS), horizontal compression (HC), anterior chamber depth (ACD), and white-to-white distance (WTW).
- Lens thickness (LT) and crystalline lens rise (CLR) negatively impacted vault, while female sex was associated with higher vaults.
Conclusions:
- A robust, AI-driven ICL vault prediction model was successfully constructed.
- The model exhibits low parameter dependency and high interpretability, offering valuable insights into vault determinants.
