Evaluating the Predictive Accuracy of an AI-Based Tool for Postoperative Vault Estimation in Phakic Intraocular Lens Implantation

Introduction: Phakic intraocular lenses are widely used for refractive error correction, with the EVO ICL delivering excellent visual outcomes. Achieving an optimal postoperative vault is critical to minimize complications. The purpose of this study was to evaluate the predictive accuracy of an AI-based tool that integrates high-resolution ultrasound biomicroscopy (UBM) imaging with biometric data, for estimating postoperative vault in myopic patients. Settings: The study was performed at four centers: Instituto Zaldivar (Argentina), Wellington Eye Clinic (Ireland), Medipolis Eye Center (Belgium), and Asian Eye Institute (Philippines). Methods: In this retrospective, multicenter study, 347 eyes from 228 myopic patients (mean age 31.3 ± 7.7 years) underwent ICL implantation. Preoperative biometric parameters and UBM imaging were utilized to generate vault predictions using the AI-based tool. Predicted vault values were compared with clinical measurements obtained at 1 day and 1 month postoperatively. Statistical analyses, including Spearman correlation and multivariable linear regression, were conducted to assess the agreement between predicted and measured vaults and to identify significant predictive factors. Results: At 1 day postoperatively, the mean clinical vault was 520.97 ± 178.73 μm versus a predicted vault of 508.16 ± 163.00 μm, with a mean signed difference of − 12.81 μm (r²=0.621, p< 0.001). Subgroup analyses across the four centers demonstrated stable predictions, with no significant inter-center differences in either clinical or predicted vault measurements (p> 0.05). Multivariable regression identified ARise and spherical power as significant predictors of vault discrepancy, with uniform effects across diverse populations. Conclusion: The ICLGuru™ reliably predicts postoperative vault with clinically acceptable accuracy. These findings underscore the generalizability and reliable performance of the AI-based tool across varied clinical settings. Its integration into preoperative planning may enhance ICL sizing and reduce complications in myopic patients. Plain Language Summary: Phakic intraocular Collamer lenses (ICLs) are special lenses implanted in the eye to correct vision problems like nearsightedness. Choosing the right size is crucial to ensure the lens fits well and does not cause complications. Doctors currently use different methods to estimate the best lens size, but these are not always perfect. This study tested a new artificial intelligence (AI) tool that predicts how much space (or “vault”) will be left between the implanted lens and the eye’s natural structures after surgery. The aim is to know if this tool could accurately estimate the vault before surgery, helping doctors choose the right lens size. To find out, 347 eyes from 228 patients who had ICL surgery in four different eye clinics around the world were analyzed. The AI tool used high-resolution ultrasound images and biometric data (measurements of the eye) to predict the vault. These predictions were then compared with actual vault measurements taken after surgery. The results showed that the AI-based tool was highly accurate, with only a small difference between predicted and actual vault values. The tool worked well across different patient groups and clinic locations. The study also identified factors, such as the shape of the eye, that slightly affected prediction accuracy. These findings suggest that AI can help eye surgeons make better decisions about lens sizing, potentially reducing the risk of complications and improving surgical outcomes for patients.