Related Experiment Video
Updated: Mar 12, 2026
10:42
A Postoperative Evaluation Guideline for Computer-Assisted Reconstruction of the Mandible
Published on: January 28, 2020
7.1K
Generative Artificial Intelligence for Postoperative Parameters Prediction in Implantable Collamer Lens Surgery
Yinglin Zhang1, Ruiling Xi1, Lingxi Zeng1
1Research Institute of Trustworthy Autonomous Systems and Department of Computer Science and Engineering, Southern University of Science and Technology, Shenzhen, 518055, China.
Journal of Cataract and Refractive Surgery
|March 10, 2026
Summary
Generative artificial intelligence accurately predicts key outcomes after implantable collamer lens (ICL) surgery using preoperative images. This AI tool aids in anticipating postoperative parameters, improving surgical planning and patient results.
Area of Science:
- Ophthalmology
- Artificial Intelligence
- Medical Imaging
Background:
- Implantable collamer lens (ICL) surgery is a common refractive procedure.
- Accurate prediction of postoperative outcomes is crucial for successful ICL implantation.
- Preoperative anterior segment optical coherence tomography (AS-OCT) provides valuable anatomical information.
Purpose of the Study:
- To develop and validate a generative artificial intelligence (AI) model for predicting multiple postoperative parameters following ICL surgery.
- To utilize preoperative AS-OCT images as input for the AI model.
- To assess the accuracy and reliability of AI-driven predictions for key surgical outcomes.
Main Methods:
- A retrospective study involving 1010 patients (1585 eyes) for horizontal ICL and 86 patients (86 eyes) for vertical ICL implantation.
- Development of a Generative Adversarial Network (ICL-GAN) to predict postoperative structures from preoperative AS-OCT.
- Measurement of postoperative parameters (vault, AOD500, TIA500) from predicted structures and evaluation of prediction error (MAE, RMSE).
Main Results:
- ICL-GAN demonstrated strong correlations between predicted and achieved vault values for horizontal ICL implantation across different lens sizes (r=0.659 to 0.799, p<0.01).
- The AI model achieved minimal prediction errors comparable to established formulas (NK, KS) for vault prediction.
- Good correlation and agreement were observed for predicted AOD500 and TIA500 values, with superior performance on vertical implantation data.
Conclusions:
- Generative artificial intelligence, specifically ICL-GAN, effectively predicts multiple postoperative parameters after ICL surgery.
- The AI model shows significant potential for enhancing surgical planning and improving patient outcomes in refractive lens surgery.
- Preoperative AS-OCT imaging combined with AI offers a promising approach for personalized ICL surgery.
