Dataset Information

Accuracy Improvement of IOL Power Prediction for Highly Myopic Eyes With an XGBoost Machine Learning-Based Calculator

ABSTRACT: Purpose: To develop a machine learning-based calculator to improve the accuracy of IOL power predictions for highly myopic eyes. Methods: Data of 1,450 highly myopic eyes from 1,450 patients who had cataract surgeries at our hospital were used as internal dataset (train and validate). Another 114 highly myopic eyes from other hospitals were used as external test dataset. A new calculator was developed using XGBoost regression model based on features including demographics, biometrics, IOL powers, A constants, and the predicted refractions by Barrett Universal II (BUII) formula. The accuracies were compared between our calculator and BUII formula, and axial length (AL) subgroup analysis (26.0–28.0, 28.0–30.0, or ?30.0 mm) was further conducted. Results: The median absolute errors (MedAEs) and median squared errors (MedSEs) were lower with the XGBoost calculator (internal: 0.25 D and 0.06 D2; external: 0.29 D and 0.09 D2) vs. the BUII formula (all P ? 0.001). The mean absolute errors and were 0.33 ± 0.28 vs. 0.45 ± 0.31 (internal), and 0.35 ± 0.24 vs. 0.43 ± 0.29 D (external). The mean squared errors were 0.19 ± 0.32 vs. 0.30 ± 0.36 (internal), and 0.18 ± 0.21 vs. 0.27 ± 0.29 D2 (external). The percentages of eyes within ±0.25 D of the prediction errors were significantly greater with the XGBoost calculator (internal: 49.66 vs. 29.66%; external: 78.28 vs. 60.34%; both P < 0.05). The same trend was in MedAEs and MedSEs in all subgroups (internal) and in AL ?30.0 mm subgroup (external) (all P < 0.001). Conclusions: The new XGBoost calculator showed promising accuracy for highly or extremely myopic eyes.

SUBMITTER: Wei L

PROVIDER: S-EPMC7793738 | biostudies-literature | 2020 Jan

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:BackgroundPrevalence of high myopia is continuously increasing, thus, patients affected with staphyloma are abundant worldwide. Assessment of the quality of vision in these patients is mandatory for a proper clinical counselling, specially when undergoing surgical procedures that require intraocular lenses implantation. Thus, the purpose of the study was to assess monochromatic higher order aberrations (HOAs) in highly myopic eyes with staphyloma with or without a dome-shaped macula.MethodsParticipants underwent spectral-domain optical coherence tomography, ocular axial biometry, dual Scheimpflug photography and integrated Placido disk topography, and Hartmann-Shack wavefront analysis. Five groups were evaluated: a low-moderate myopia control group (< 6.00 diopters, n = 31) and four high myopia (≥6.00 diopters) groups: eyes without staphyloma (n = 18), eyes with inferior staphyloma (n = 14), eyes with posterior staphyloma without dome-shaped macula (n = 15) and eyes with posterior staphyloma with dome-shaped macula (n = 17). Subsequently, two new groups (including all participants) were created to assess differences between myopia with and without staphyloma. One-way analysis of covariance was performed using age and lens densitometry as covariates.ResultsStatistically significant (p ≤ 0.05) differences in anterior corneal fourth-order HOAs were observed between the low-moderate myopia and no-dome-shaped macula (Mean: 0.16 μm) and dome-shaped macula posterior staphyloma groups (Mean: 0.12 μm) in younger patients (≤45 years old). The same groups also showed (p ≤ 0.05) significant differences for anterior corneal primary spherical aberration (Mean: 0.19 and 0.13 μm, respectively). In addition, anterior corneal tetrafoil was significantly higher (p = 0.04) in dome-shaped macula compared to no-dome-shaped macula (Mean: 0.18 vs 0.06 μm, respectively). When all participants were grouped together, significantly lower mean anterior corneal primary spherical aberration (0.15 μm vs. 0.27 μm, p = 0.004) and higher internal primary spherical aberration (0.04 μm vs. -0.06 μm, p = 0.04) was observed in staphyloma compared to no-staphyloma myopic patients.ConclusionsEyes with high myopia and staphyloma have less positive anterior corneal primary spherical aberration and less negative internal primary spherical aberration, suggesting that the anterior corneal surface tends to mimic in a specular fashion the posterior pole profile. This corneal behaviour appears to change in patients older than 45 years.

Project description:AimTo evaluate the accuracy of the raytracing method for the calculation of intraocular lens (IOL) power in myopic eyes after small incision extraction of the lenticule (SMILE).MethodsRetrospective study. All patients undergoing surgery for myopic SMILE between May 1, 2020, and December 31, 2020, with Scheimpflug tomography optical biometry were eligible for inclusion. Manifest refraction was performed before and 6 months after refractive surgery. One eye from each patient was included in the final analysis. A theoretical model was invited to predict the accuracy of multiple methods of lens power calculation by comparing the IOL-induced refractive error at the corneal plane (IOL-Dif) and the SMILE-induced change of spherical equivalent (SMILE-Dif) before and after SMILE surgery. The prediction error (PE) was calculated as the difference between SMILE-Dif-IOL-Dif. IOL power calculations were performed using raytracing (Olsen Raytracing, Pentacam AXL, software version 1.22r05, Wetzlar, Germany) and other formulae with historical data (Barrett True-K, Double-K SRK/T, Masket, Modified Masket) and without historical data (Barrett True-K no history, Haigis-L, Hill Potvin Shammas PM, Shammas-PL) for the same IOL power and model. In addition, subgroup analysis was performed in different anterior chamber depths, axial lengths, back-to-front corneal radius ratio, keratometry, lens thickness, and preoperative spherical equivalents.ResultsA total of 70 eyes of 70 patients were analyzed. The raytracing method had the smallest mean absolute PE (0.26 ± 0.24 D) and median absolute PE (0.16 D), and also had the largest percentage of eyes within a PE of ± 0.25 D (64.3%), ± 0.50 D (81.4%), ± 0.75 D (95.7%), and ± 1.00 D (100.0%). The raytracing method was significantly better than Double-K SRK/T, Haigis, Haigis-L, and Shammas-PL formulae in postoperative refraction prediction (all p < 0.001), but not better than the following formulae: Barrett True-K (p = 0.314), Barrett True-K no history (p = 0.163), Masket (p = 1.0), Modified Masket (p = 0.806), and Hill Potvin Shammas PM (p = 0.286). Subgroup analysis showed that refractive outcomes exhibited no statistically significant differences in the raytracing method (all p < 0.05).ConclusionRaytracing was the most accurate method in predicting target refraction and had a good consistency in calculating IOL power for myopic eyes after SMILE.

Dataset Information

Accuracy Improvement of IOL Power Prediction for Highly Myopic Eyes With an XGBoost Machine Learning-Based Calculator

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets