Artificial Intelligence Prediction of Postoperative Rotation Stability After Toric Implantable Collamer Lens
Yinjie Jiang1,2,3,4, Xun Chen1,2,3,4, Mingrui Cheng1,2,3,4
1Fudan University Eye Ear Nose and Throat Hospital, Shanghai, China.
Translational Vision Science & Technology
|November 13, 2025
Summary
Artificial intelligence models accurately predict toric implantable collamer lens rotation and its impact on vision. This aids ophthalmologists in anticipating outcomes and managing potential complications after surgery.
Area of Science:
- Ophthalmology
- Artificial Intelligence
- Medical Imaging
Background:
- Toric implantable collamer lens (TICL) surgery is a common procedure for correcting refractive errors.
- Postoperative rotation of TICL can lead to suboptimal refractive outcomes and visual disturbances.
- Predicting TICL rotation and its consequences is crucial for optimizing surgical results.
Purpose of the Study:
- To develop and evaluate artificial intelligence (AI) models for predicting postoperative TICL rotation.
- To assess the impact of TICL rotation on refraction and visual acuity.
- To identify critical rotation thresholds that affect visual outcomes and necessitate intervention.
Main Methods:
- Utilized data from 642 eyes of 371 patients who underwent TICL surgery.
- Employed regression models to predict rotation degrees and classification models for rotation-related complications.
- Evaluated model performance using metrics such as RMSE, R², accuracy, and AUC; conducted subgroup analyses for different astigmatism levels.
Main Results:
- The Tabular prior-data fitted network (TabPFN) demonstrated superior accuracy in predicting TICL rotation (RMSE: 10.672 ± 5.880).
- TabPFN achieved high accuracy (0.906-0.981) in predicting rotation-related complications, including a near-perfect precision (0.990 ± 0.006) for realignment surgery prediction.
- Identified specific rotation cutoff values impacting astigmatism and visual acuity across different astigmatism groups, with AUCs ranging from 0.65 to 0.93.
Conclusions:
- AI models, particularly TabPFN, are effective tools for predicting postoperative TICL rotation and associated visual outcomes.
- These AI models provide valuable predictive insights, assisting ophthalmologists in surgical planning and patient management.
- The study highlights the successful integration of AI in ophthalmology for enhancing the precision and predictability of refractive surgery.
More Related Videos
05:46Author Spotlight: Advancements in Refractive Surgical Correction for Presbyopia and Exploring Postoperative Visual Acuity
Published on: September 20, 2024
754
05:14Comparison of Agreement and Accuracy using Binocular Wavefront Optometer with Autorefractor and Phoropter
Published on: September 16, 2025
548