PyGlaucoMetrics: A Stacked Weight-Based Machine Learning Approach for Glaucoma Detection Using Visual Field Data

Medicina, Journal Year: 2025, Volume and Issue: 61(3), P. 541 - 541

Published: March 20, 2025

Background and Objectives: Glaucoma (GL) classification is crucial for early diagnosis treatment, yet relying solely on stand-alone models or International Classification of Diseases (ICD) codes insufficient due to limited predictive power inconsistencies in clinical labeling. This study aims improve GL using stacked weight-based machine learning models. Materials Methods: We analyzed a subset 33,636 participants (58% female) with 340,444 visual fields (VFs) from the Mass Eye Ear (MEE) dataset. Five clinically relevant detection (LoGTS, UKGTS, Kang, HAP2_part1, Foster) were selected serve as base Two multi-layer perceptron (MLP) trained 52 total deviation (TD) pattern (PD) values Humphrey field analyzer (HFA) 24-2 VF tests, along four variables (age, gender, follow-up time, race) extract model weights. These weights then utilized train three meta-learners, including logistic regression (LR), extreme gradient boosting (XGB), MLP, classify cases non-GL. Results: The MLP meta-learner achieved highest performance, an accuracy 96.43%, F-score 96.01%, AUC 97.96%, while also demonstrating lowest prediction uncertainty (0.08 ± 0.13). XGB followed 92.86% accuracy, 92.31% F-score, 96.10% AUC. LR had 89.29% 86.96% 94.81% AUC, well (0.58 0.07). Permutation importance analysis revealed that superior temporal sector was most influential feature, scores 0.08 Kang’s 0.04 HAP2_ part1 Among variables, age strongest contributor (score = 0.3). Conclusions: outperformed classification, achieving improvement 8.92% over best-performing (LoGTS 87.51%), offering valuable tool automated glaucoma detection.

Language: Английский

---

Stratified Multisource Optical Coherence Tomography Integration and Cross-Pathology Validation Framework for Automated Retinal Diagnostics

Applied Sciences, Journal Year: 2025, Volume and Issue: 15(9), P. 4985 - 4985

Published: April 30, 2025

This study presents a clinical utility-driven machine learning framework for retinal Optical Coherence Tomography classification, addressing challenges posed by manual interpretation variability and dataset heterogeneity. The methodology integrates biomimetic data partitioning, deep biomarker extraction via pretrained VGG16 networks, automated model selection optimized decision-making. Stratified curation preserved pathological distributions across training, validation, testing subsets, while SMOTE optimization mitigated class imbalance. Cross-pathology evaluated generalizability on anatomically distinct conditions excluded from assessing the framework’s robustness to unseen pathologies. Clinical utility metrics prioritized alignment with ophthalmological imperatives, emphasizing negative predictive value minimize false negatives enhance diagnostic reliability. advances AI-driven diagnostics harmonizing computational performance patient-centered outcomes, enabling standardized disease detection diverse datasets through robust feature generalization.

Language: Английский

---

Translating the machine; An assessment of clinician understanding of ophthalmological artificial intelligence outputs

International Journal of Medical Informatics, Journal Year: 2025, Volume and Issue: 201, P. 105958 - 105958

Published: May 6, 2025

Language: Английский

---

Multimodal AI diagnostic system for neuromyelitis optica based on ultrawide-field fundus photography

Frontiers in Medicine, Journal Year: 2025, Volume and Issue: 12

Published: May 7, 2025

While deep learning (DL) has demonstrated significant utility in ocular diseases, no clinically validated algorithm currently exists for diagnosing neuromyelitis optica (NMO). This study aimed to develop a proof-of-concept multimodal artificial intelligence (AI) diagnostic model that synergistically integrates ultrawide field fundus photographs (UWFs) with clinical examination data predicting the onset and stage of suspected NMO. The utilized UWFs 330 eyes from 285 NMO patients 1,288 770 non-NMO participants, along reports, an AI or performance was evaluated based on area under receiver operating characteristic curve (AUC), sensitivity, specificity. achieved AUC 0.9923, maximum Youden index 0.9389, sensitivity 97.0% specificity 96.9% prevalence test set. Our demonstrates feasibility DL algorithms

Language: Английский

---