Experimental eye research · 2025
The utility of artificial intelligence in characterization and detecting causes of macular edema: A spectral-domain OCT-based algorithm study.
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.
Alzu'bi Amal, Momany Sondos, Aleshawi Abdelwahab, Tashtoush Mais, Al-Dwairi Rami
Аннотация
BACKGROUND: Macular Edema (ME), a prevalent cause of vision loss, can arise from various retinal conditions, most notably diabetic macular edema (DME) and age-related macular degeneration (AMD). Accurate and timely differentiation among these causes is necessary for appropriate treatment; however, it remains a diagnostic challenge. This research addresses the gap in automated ME classification by developing and evaluating a deep learning framework capable of distinguishing between DME, AMD, and normal retinal conditions using optical coherence tomography (OCT) images.
METHODS: A retrospective dataset comprising 1040 OCT images from King Abdullah University Hospital (KAUH) was used in conjunction with a public dataset for benchmarking. The dataset was divided into annotated and non-annotated images, with preprocessing, augmentation, and simulated segmentation applied to improve the model performance. We benchmarked and evaluated three pretrained convolutional neural networks-ResNet152, InceptionV3, and MobileNetV2.
RESULTS: Among the models, InceptionV3 and ResNet152 achieved the highest accuracies (95 %-98 %) across both datasets. MobileNetV2, on the other hand, showed moderate accuracy on the KAUH dataset (89 %) but exhibited strong performance on the public dataset (97 %). Explainable AI (XAI) techniques, specifically Grad-CAM, were applied to visualize the model predictions, and the outcomes were manually validated against annotated data to assess interpretability.
CONCLUSIONS: The findings support the integration of a robust CNN architecture and XAI techniques to enhance diagnostic precision and aid clinical decision-making in ophthalmology.
METHODS: A retrospective dataset comprising 1040 OCT images from King Abdullah University Hospital (KAUH) was used in conjunction with a public dataset for benchmarking. The dataset was divided into annotated and non-annotated images, with preprocessing, augmentation, and simulated segmentation applied to improve the model performance. We benchmarked and evaluated three pretrained convolutional neural networks-ResNet152, InceptionV3, and MobileNetV2.
RESULTS: Among the models, InceptionV3 and ResNet152 achieved the highest accuracies (95 %-98 %) across both datasets. MobileNetV2, on the other hand, showed moderate accuracy on the KAUH dataset (89 %) but exhibited strong performance on the public dataset (97 %). Explainable AI (XAI) techniques, specifically Grad-CAM, were applied to visualize the model predictions, and the outcomes were manually validated against annotated data to assess interpretability.
CONCLUSIONS: The findings support the integration of a robust CNN architecture and XAI techniques to enhance diagnostic precision and aid clinical decision-making in ophthalmology.