Background and Objectives: Glaucoma (GL) classification is crucial for early diagnosis and treatment, yet relying solely on stand-alone models or International Classification of Diseases (ICD) codes is insufficient due to limited predictive power and inconsistencies in clinical labeling. This study aims to improve GL classification using stacked weight-based machine learning models. Materials and Methods: We analyzed a subset of 33,636 participants (58% female) with 340,444 visual fields (VFs) from the Mass Eye and Ear (MEE) dataset. Five clinically relevant GL detection models (LoGTS, UKGTS, Kang, HAP2_part1, and Foster) were selected to serve as base models. Two multi-layer perceptron (MLP) models were trained using 52 total deviation (TD) and pattern deviation (PD) values from Humphrey field analyzer (HFA) 24-2 VF tests, along with four clinical variables (age, gender, follow-up time, and race) to extract model weights. These weights were then utilized to train three meta-learners, including logistic regression (LR), extreme gradient boosting (XGB), and MLP, to classify cases as GL or non-GL. Results: The MLP meta-learner achieved the highest performance, with an accuracy of 96.43%, an F-score of 96.01%, and an AUC of 97.96%, while also demonstrating the lowest prediction uncertainty (0.08 ± 0.13). XGB followed with 92.86% accuracy, a 92.31% F-score, and a 96.10% AUC. LR had the lowest performance, with 89.29% accuracy, an 86.96% F-score, and a 94.81% AUC, as well as the highest uncertainty (0.58 ± 0.07). Permutation importance analysis revealed that the superior temporal sector was the most influential VF feature, with importance scores of 0.08 in Kang’s and 0.04 in HAP2_part1 models. Among clinical variables, age was the strongest contributor (score = 0.3). Conclusions: The meta-learner outperformed stand-alone models in GL classification, achieving an accuracy improvement of 8.92% over the best-performing stand-alone model (LoGTS with 87.51%), offering a valuable tool for automated glaucoma detection.

Background and Objectives: Glaucoma (GL) classification is crucial for early diagnosis and treatment, yet relying solely on stand-alone models or International Classification of Diseases (ICD) codes is insufficient due to limited predictive power and inconsistencies in clinical labeling. This study aims to improve GL classification by stacking weights using machine learning-based models. Materials and Methods: We analyzed a subset of 33,636 participants (58% female) with 340,444 visual fields (VFs) from the Mass Eye and Ear (MEE) dataset. Five clinically relevant GL detection models (LoGTS, UKGTS, Kang, HAP2_part1, and Foster) were selected to serve as base models. Two Multi-Layer Perceptron (MLP) models were trained using 52 total deviation (TD) and pattern deviation (PD) values from Humphrey Field Analyzer (HFA) 24-2 VF tests, along with four clinical variables (age, gender, follow-up time, and race) to extract model weights. These weights were then utilized to train three meta-learners including Logistic Regression (LR), Extreme Gradient Boosting (XGB), and MLP to classify cases as GL or non-GL. Results: The MLP meta-learner achieved the highest performance, with an accuracy of 96.43%, an F-score of 96.01%, and an AUC of 97.96%, while also demonstrating the lowest prediction uncertainty (0.08 ± 0.13). XGB followed with 92.86% accuracy, a 92.31% F-score, and a 96.10% AUC. LR had the lowest performance, with 89.29% accuracy, an 86.96% F-score, and a 94.81% AUC, as well as the highest uncertainty (0.58 ± 0.07). Permutation importance analysis revealed that the superior temporal sector was the most influential VF feature, with importance scores of 0.08 in Kang’s and 0.04 in HAP2_ part1 models. Among clinical variables, age was the strongest contributor (score= 0.3). Conclusion: The meta-learning approach outperformed stand-alone models in GL classification, offering a valuable tool for automated glaucoma assessment.

Purpose: The aim of this study was to compare preoperative determinants of procedure selection and long-term outcomes between anterior lamellar keratoplasty (ALK) and penetrating keratoplasty (PKP) in patients with keratoconus in the United States. Methods: A total of 10,405 eyes from patients with keratoconus undergoing corneal transplantation were included from the Intelligent Research in Sight Registry, comprising 1005 eyes that received ALK (mean age 40.0 ±15.5 years) and 9400 eyes that received PKP (mean age 44.3 ±16.4 years) between January 1, 2013, and December 31, 2020. Univariate and multivariable logistic regression analyses were performed to identify demographic and clinical factors associated with the selection of ALK versus PKP. Results: In multivariable analysis, older age (50–59 years: odds ratio [OR]: 0.05; 95% confidence interval [CI]: 0.00–0.88; ≥70 years: OR: 0.01; 95% CI: 0.00–0.44), male sex (OR: 0.28; 95% CI: 0.09–0.90), and Medicaid insurance (OR: 0.11; 95% CI: 0.02–0.55) were significantly associated with lower odds of undergoing ALK. Graft failures (as measured by the incidence of regrafts) and rejections were significantly less frequent in ALK at 90 days (failure: 3.3% vs. 12.6%, P < 0.001; rejection: 1.9% vs. 5.8%, P < 0.001) and at 1 year (failure: 5.4% vs. 12.3%, P < 0.001; rejection: 6.5% vs. 11.7%, P < 0.001), but not at 3 and 5 years. Conclusions: In the United States, younger patients are more likely to receive ALK. Other socioeconomic factors such as ethnicity and insurance coverage may also play an important role. However, our data suggest that ALK may not result in improved postoperative outcomes compared with PKP over 5 years.