Artificial intelligence (AI) may help predict clinical outcomes of implantable collamer lens (ICL) surgery, according to findings published in the British Journal of Ophthalmology.
Posterior chamber phakic intraocular lens implants — including non-toric ICLs — are increasing in popularity, according to the report. It has become one of the mainstream surgeries for myopia correction, particularly for patients with high myopia. However, improper ICL size is a frequently reported complication of implantation.
To improve outcomes, researchers examined 3536 patients implanted with ICLs (6297 eyes). Vault values were measured, and the team investigated the importance of vault and input parameters.
In the prediction of the vault, the Random Forest model provided the best results in the regression model (R2=0.315), followed by the Gradient Boosting model (R2=0.291) and XGBoost model (R2=0.285). The maximum classification accuracy is 0.828 in Random Forest, and the mean area under the curve (AUC) is 0.765. Investigators found that Random Forest predicts the ICL size with 82.2% accuracy. Gradient Boosting and XGBoost provide 81.5% and 81.8% accuracy, respectively.
“AI is applicable for vault prediction and ICL sizing. Random Forests, Gradient Boosting and XGBoost are the most preferred [machine learning] models,” researchers report.
The study’ authors noted some limitations to their study, including that as a cross-sectional study, it doesn’t assess the vault changes over time. Additional limitations include the relatively small sample size of the patients implanted with extremely small (12.1 mm) or large (13.7 mm) ICLs, that only a single source of data is used, and that ICL vaults were predicted mainly using white-to-white values.
Shen Y, Wang L, Jian W, et al. Big-data and artificial-intelligence-assisted vault prediction and EVO-ICL size selection for myopia correction. Br J Ophthalmol. Published Online September 6, 2021. doi:10.1136/bjophthalmol-2021-319618