Aug 2026· Clinical Ophthalmology· Vol 20· 0 citations· 31 references
Medicine
TL;DR
ICLGuru demonstrated useful postoperative vault predictions with performance comparable to existing machine learning and UBM-based nomograms, though it appeared to overestimate longer term vault in the authors' cohort, and may be valuable in identifying eyes at risk for excessive vault.
Abstract
Purpose Our study evaluated a cohort of patients who previously underwent EVO Implantable Collamer Lens (ICL) implantation to assess the retrospective performance of the ICLGuru™ artificial intelligence (AI) sizing tool in predicting postoperative vault. Additionally, biometric factors associated with vault prediction error were identified. Methods This retrospective single-center study included 104 eyes from 56 patients who underwent EVO V4c or V5 ICL implantation between May 2023 and June 2025. Preoperative ultrasound biomicroscopy (UBM) images, from patients who had already undergone ICL surgery, were retrospectively analyzed using ICLGuru to generate predicted postoperative vaults. Actual postoperative vault was measured at 1-, 3-, or 6-month follow-up visits, with 3-month measurements preferentially included for analysis. Mixed-effects regression models were used to evaluate associations between biometric variables and vault prediction error. Results Mean absolute error was 135 ± 110 µm, indicating a tendency for ICLGuru to overestimate postoperative vault (p < 0.001). Among 16 eyes with actual postoperative overvault, ICLGuru retrospectively identified 15 cases in which a smaller lens size was predicted to achieve target vault, suggesting these outcomes may have been flagged. Out of 7 eyes with actual undervault, only two cases may have been flagged with use of ICLGuru, indicating that the tool is not as useful in preventing undervault. Mixed-effects regression demonstrated that greater iridocorneal angle and higher aRISE were significantly associated with lower vault prediction error (p < 0.05). Conclusion ICLGuru demonstrated useful postoperative vault predictions with performance comparable to existing machine learning and UBM-based nomograms, though it appeared to overestimate longer term vault in our cohort. The tool may be valuable in identifying eyes at risk for excessive vault. aRISE and iridocorneal angle were significant predictors of vault prediction error, supporting their importance in future ICL sizing models.
Background To comprehensively explore the biometry related to suboptimal vault (including low and high vault) after implantable collamer lens (ICL) implantation and the risk factors associated with ICL. Methods This retrospective case-control study enrolled 180 patients with myopia and astigmatism who underwent EVO-V4c ICL implantation at Peking Union Medical College Hospital between January 2023 and December 2024. The normal vault group (NVG) was randomly selected from eligible patients during the same period to ensure balanced group sizes and reduce selection bias. Patients were divided into three groups based on the postoperative vault: low vault group (LVG, <250 μm, n = 36), high vault group (HVG, >750 μm, n = 72), and normal vault group (NVG, 250–750 μm, n = 72). Biometric parameters were collected before and after surgery for all patients, including anterior chamber depth (ACD), white-to-white (WTW), angle-to-angle (ATA), crystalline lens rise (CLR), axial length (AL), and spherical equivalent (SE). Multivariate analysis of variance (MANOVA), Pearson correlation analysis, multiple linear regression, and logistic regression analysis were employed to assess the relationship between each parameter and postoperative vault outcomes. Results The results indicated significant differences in age, ACD, WTW, ATA, AL, and CLR among the three groups of patients (P < 0.05). Multivariate linear regression analysis revealed that ACD (P < 0.05), CLR (P < 0.001), and ATA (P = 0.018) were independent factors affecting postoperative vault. Logistic regression analysis showed that higher CLR (OR = 1.008, P = 0.013) was an independent risk factor for low vault, while deeper ACD (OR = 1.477, P = 0.030) and more negative ICL spherical equivalent (ICLSE, OR = 0.523, P = 0.034) were independent risk factors for high vault. Conclusions Low vault is independently associated with greater CLR, whereas high vault is independently linked to deeper ACD and higher SE. Age, WTW, ATA, and AL also demonstrate significant but varying associations with vault type. These findings underscore the importance of integrating CLR and ATA measurements into routine preoperative assessment. Employing a more individualized sizing strategy—particularly in patients with deeper ACD or pronounced CLR—could enhance vault predictability and improve postoperative safety.
Xi Yang, Shan Yang, Zhikun Yang et al.· Frontiers in Medicine· 0 citations
Purpose To evaluate and compare the prediction accuracy of toric IOL power calculation formulas available on the ESCRS online calculator, the Abulafia-Koch combined with Hill-RBF and Castrop formulae, and assess their effectiveness in predicting actual postoperative refractive outcomes. Patients and Methods The preoperative and 4–5 week postoperative records of 131 patients who underwent phacoemulsification and implantation of enVista toric (MX60T) IOL were reviewed retrospectively. The IOL power calculation formulas available on the ESCRS online calculator (Kane, EVO, Hoffer QST, and Barrett), Abulafia-Koch combined with Hill-RBF and Castrop formulae, were used to obtain predicted postoperative refraction for the implanted IOL. Spherical equivalent and residual astigmatism prediction errors (SEQ-PE and RA-PE) were compared statistically using Eyetemis. Results were expressed as trueness, accuracy, and precision. Results Mean corneal astigmatism decreased from 2.47±0.95 D preoperatively to 0.34±0.28 D postoperatively; 88.6% of eyes achieved cylinder ≤0.50 D. For SEQ-PE, trueness was comparable for Abulafia-Koch combined with Hill-RBF, Castrop, EVO, and Barrett, with trimmed means ranging between −0.02 to 0.02 D. Precision was similar for EVO, Kane, and Barrett (0.16–0.18 D) and lower for Abulafia–Koch combined with Hill-RBF and Hoffer QST formulas. Accuracy was comparable among Barrett, Castrop, EVO, and Kane, with Barrett, EVO, and Kane formulas achieving the highest proportion of eyes (>93%) within ±0.50 D of prediction error. For RA-PE, the trueness and precision of Abulafia-Koch combined with Hill-RBF, Hoffer QST, EVO, Kane, and Barrett formulas were high and comparable. Trueness ranged from 0.08 to 0.14 D and precision from 0.34 D to 0.37 D. All formulas except Castrop showed good astigmatic prediction accuracy, with more than two-thirds of eyes achieving absolute RA-PE within 0.50 D. Conclusion Among the evaluated formulas, the EVO, Barrett, and Kane formulas showed high refractive and astigmatic predictability, supporting their reliability for toric IOL power calculation.
F. Castro-Alonso, Rubén Hernández Vián, Marta Jiménez-García et al.· Clinical Ophthalmology· 0 citations
This study successfully developed and validated an explainable RF-based machine learning model for the prediction of postoperative ACD in highly myopic cataract patients, which supports more reliable IOL power calculation and offers a practical tool for optimizing surgical planning in highly myopic eyes.
Yuyang Yang, Hao Cui, Jiajia Gao et al.· Frontiers in Cell and Develo...· 0 citations
Structured preoperative variables can partially reproduce single-center clinician-selected refractive procedure patterns but do not establish optimal surgical recommendation or external generalizability.
Yinhao Li, Gang Li, Chuanyun Xu et al.· BMC Ophthalmology· 0 citations
Accurate preoperative identification of patients with Fuchs endothelial corneal dystrophy (FECD) who are at risk of requiring endothelial keratoplasty (EK) after cataract surgery remains challenging. The SUbClinical Corneal Edema Scheimpflug Study (SUCCESS) score was previously developed to estimate this risk using Scheimpflug tomography. We aimed to prospectively and externally validate the SUCCESS score in an independent multicenter cohort and determine whether the incorporation of corneal densitometry improves predictive performance and clinical utility. In this prospective multicenter cohort study, 207 eyes from 177 patients with FECD grade ≥ 2 and visually significant cataract were enrolled across six tertiary hospitals in Spain (June 2020 to October 2024), with follow-up through May 2025. After predefined exclusions, 178 eyes (149 patients) were analyzed. The original SUCCESS score was used without recalibration. An extended model incorporated peak mean corneal densitometry within the central 3-mm zone. Time-to-event analyses were performed using Cox proportional hazards models. Discrimination, calibration, reclassification, and clinical utility were assessed using Harrell’s C-index, Brier score, net reclassification index (NRI), integrated discrimination improvement (IDI), and decision curve analysis. Over a median follow-up of 37.0 months (interquartile range [IQR], 10.1–45.9), 38 eyes (21%) met the criteria for EK, with 82% of events occurring within 6 months after phacoemulsification. The original SUCCESS score demonstrated good discrimination (C-index = 0.80; 95% confidence interval [CI], 0.73–0.86) with slight underestimation of absolute risk (predicted 16.4% vs. observed 21.6%). Incorporation of corneal densitometry improved discrimination (C-index = 0.85; 95% CI, 0.79–0.90; ΔC = + 0.05; P = 0.0047), reduced prediction error (ΔBrier = − 0.019; P < 0.001), and enhanced risk reclassification (NRI = 0.39; 95% CI, 0.14–0.64; P = 0.002), including correct upward reclassification of 26% of eyes requiring EK. The extended model provided greater net benefit at relevant decision thresholds (25%–50%). In this multicenter external validation study, the SUCCESS score demonstrated a robust performance in predicting postoperative EK in FECD. The addition of corneal densitometry provided incremental improvements in discrimination and clinically meaningful risk stratification, particularly in intermediate-risk cases. Standardized Scheimpflug-based risk assessment may support individualized surgical planning and referral decisions.
F. Arnalich-Montiel, Martha Stokking, N. Sáenz-Madrazo et al.· Eye and Vision· 0 citations
The relationship between preoperative biometrics and postoperative equivalent power appears predominantly linear, Though larger datasets may enhance machine-learning performance, though larger datasets may enhance machine-learning performance.
Amanda Pan, K. P. Kaiser, Stefan Raidl et al.· Current Eye Research· 0 citations