Related Experiment Video
Updated: Jul 5, 2025
09:10
Digital Hybrid Model Preparation for Virtual Planning of Reconstructive Dentoalveolar Surgical Procedures
Published on: August 5, 2021
1.8K
Predicting Implantable Collamer Lens Vault Using Machine Learning Based on Various Preoperative Biometric Factors
Yu Di1, Huihui Fang2,3, Yan Luo1
1Department of Ophthalmology, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences, Beijing, China.
Translational Vision Science & Technology
|January 15, 2024
Summary
Machine learning models, particularly XGBoost, accurately predict vault size after Implantable Collamer Lens (ICL) V4c surgery. This approach offers improved prediction over traditional methods, potentially reducing complications and secondary surgeries.
Area of Science:
- Ophthalmology
- Medical Artificial Intelligence
- Biomedical Engineering
Background:
- The vault, the anterior chamber depth post-implantation, is crucial for refractive outcomes and preventing complications after Implantable Collamer Lens (ICL) V4c surgery.
- Accurate prediction of postoperative vault size is essential for optimizing ICL outcomes and minimizing risks such as pigment dispersion syndrome or angle closure.
Purpose of the Study:
- To develop and evaluate machine learning models for predicting the vault size following ICL V4c implantation.
- To compare the predictive accuracy of machine learning algorithms against the manufacturer's standard nomogram.
Main Methods:
- A retrospective study included 707 eyes undergoing ICL V4c implantation.
- Machine learning models, including Random Forest Regression (RFR), XGBoost, and Linear Regression (LR), were trained to predict vault size one week post-surgery.
- Model performance was assessed using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Symmetric Mean Absolute Percentage Error (SMAPE), and Bland-Altman analysis.
Main Results:
- XGBoost demonstrated the lowest prediction error, with MAE of 121.70 µm, RMSE of 148.87 µm, and SMAPE of 19.13%.
- Bland-Altman plots indicated that XGBoost and RFR provided more consistent predictions than LR.
- XGBoost exhibited narrower 95% limits of agreement (-307.12 to 256.59 µm) compared to RFR, suggesting superior predictive precision.
Conclusions:
- XGBoost outperformed RFR and LR in predicting postoperative vault size after ICL V4c implantation.
- Machine learning models show significant potential for accurate vault prediction, which could aid in reducing surgical complications and the need for revision surgeries.
