Graefe's archive for clinical and experimental ophthalmology = Albrecht von Graefes Archiv fur klinische und experimentelle Ophthalmologie · 2025
Effective automatic classification methods via deep learning for multi-type infectious keratitis diagnosis.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Zhang Yang, Wang Yuning, Xu Yingnan, Yang Weihua
Аннотация
BACKGROUND: Infectious keratitis (IK) is a leading cause of corneal blindness, typically caused by bacteria, fungi, viruses, or parasites. Prompt diagnosis and treatment are crucial, yet the absence of a gold standard for pathogen identification complicates timely interventions. Corneal cultures can be time-consuming and prone to false positives, highlighting the need for an automated classification system.
METHODS: From March 2018 to November 2023, 1,065 diffuse pattern slit-lamp images were collected to develop a deep learning system. Five models-EfficientNet_B0, EfficientNet_V2_S, ResNet50, Vision Transformer (ViT), and DeepIK-were trained for corneal infection classification. Key evaluation metrics included accuracy, precision, recall, F1-score, weighted Cohen's Kappa, and the Receiver Operating Characteristic (ROC) curve.
RESULTS: The EfficientNet_B0 model achieved superior performance across all metrics, with an accuracy of 75.2% (95% CI: 69.6% - 80.8%), sensitivity of 74.9% (95% CI: 69.9% - 80.3%), specificity of 93.8% (95% CI: 92.4% - 95.2%), F1-Score of 74.3% (95% CI: 68.7% - 79.8%), Kappa value of 0.689 (95% CI: 0.618-0.759), and AUC of 0.943 (95% CI: 0.920-0.962).
CONCLUSIONS: The EfficientNet_B0 model effectively identified normal eyes and four IK types, showcasing the potential of deep learning in diagnosing keratitis infections. Future enhancements with larger datasets could improve accuracy, facilitating timely treatments and better outcomes for patients.
METHODS: From March 2018 to November 2023, 1,065 diffuse pattern slit-lamp images were collected to develop a deep learning system. Five models-EfficientNet_B0, EfficientNet_V2_S, ResNet50, Vision Transformer (ViT), and DeepIK-were trained for corneal infection classification. Key evaluation metrics included accuracy, precision, recall, F1-score, weighted Cohen's Kappa, and the Receiver Operating Characteristic (ROC) curve.
RESULTS: The EfficientNet_B0 model achieved superior performance across all metrics, with an accuracy of 75.2% (95% CI: 69.6% - 80.8%), sensitivity of 74.9% (95% CI: 69.9% - 80.3%), specificity of 93.8% (95% CI: 92.4% - 95.2%), F1-Score of 74.3% (95% CI: 68.7% - 79.8%), Kappa value of 0.689 (95% CI: 0.618-0.759), and AUC of 0.943 (95% CI: 0.920-0.962).
CONCLUSIONS: The EfficientNet_B0 model effectively identified normal eyes and four IK types, showcasing the potential of deep learning in diagnosing keratitis infections. Future enhancements with larger datasets could improve accuracy, facilitating timely treatments and better outcomes for patients.