Ophthalmology science · 2025
Detecting Inflammation in Fundus Photographs Using Machine Learning.
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.
Mohammadi S Saeed, Yavari Negin, Saengsirinavin Aim-On, Hung Jia-Horung, Yoo Woong-Sun, Añover Frances Andrea, Mobasserian Azadeh, Gupta Ankur Sudhir, Akhavanrezayat Amir, Tuong Than Ngoc Trong, Or Christopher, Thng Zheng Xian, Karaca Irmak, Gill Manjot K, Nguyen Quan Dong
Аннотация
PURPOSE: To utilize a machine learning model for employing ultra-widefield fundus photograph (UWFFP) as a surrogate marker for ultra-widefield fluorescein angiography (UWFFA) in detecting posterior segment inflammation.
DESIGN: An evaluation of technology.
PARTICIPANTS: Not applicable.
METHODS: Ultra-widefield fundus photographs were extracted and grouped based on the corresponding UWFFA images, which served as the ground truth, into those with signs of inflammation and those without. Vertex AI was used to train a machine learning model to detect the presence of inflammation using the UWFFPs. Furthermore, 2 masked dual fellowship-trained vitreoretinal and uveitis specialists, 1 glaucoma specialist, and 1 comprehensive ophthalmologist independently evaluated 20 additional UWFFPs, comparing their findings with the model's results.
MAIN OUTCOME MEASURES: Area under the precision-recall curve, sensitivity, specificity, and comparative accuracy between the machine learning model and expert human graders.
RESULTS: A total of 302 UWFFPs, comprising 113 with inflammation and 189 without, were used to train a single-label image classification model. The trained model demonstrated an area under the curve of 0.943, sensitivity of 90.91%, and specificity of 84.21%. It correctly diagnosed the presence of inflammation in 95% of the extra UWFFPs, compared to the masked dual fellowship-trained vitreoretinal and uveitis specialists, glaucoma specialist, and comprehensive ophthalmologist, who achieved accuracies of 85%, 80%, 70%, and 65%, respectively.
CONCLUSIONS: The index study demonstrated that UWFFPs could be employed as a noninvasive and more accessible imaging modality for detecting posterior segment inflammation using machine learning techniques.
FINANCIAL DISCLOSURES: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.
DESIGN: An evaluation of technology.
PARTICIPANTS: Not applicable.
METHODS: Ultra-widefield fundus photographs were extracted and grouped based on the corresponding UWFFA images, which served as the ground truth, into those with signs of inflammation and those without. Vertex AI was used to train a machine learning model to detect the presence of inflammation using the UWFFPs. Furthermore, 2 masked dual fellowship-trained vitreoretinal and uveitis specialists, 1 glaucoma specialist, and 1 comprehensive ophthalmologist independently evaluated 20 additional UWFFPs, comparing their findings with the model's results.
MAIN OUTCOME MEASURES: Area under the precision-recall curve, sensitivity, specificity, and comparative accuracy between the machine learning model and expert human graders.
RESULTS: A total of 302 UWFFPs, comprising 113 with inflammation and 189 without, were used to train a single-label image classification model. The trained model demonstrated an area under the curve of 0.943, sensitivity of 90.91%, and specificity of 84.21%. It correctly diagnosed the presence of inflammation in 95% of the extra UWFFPs, compared to the masked dual fellowship-trained vitreoretinal and uveitis specialists, glaucoma specialist, and comprehensive ophthalmologist, who achieved accuracies of 85%, 80%, 70%, and 65%, respectively.
CONCLUSIONS: The index study demonstrated that UWFFPs could be employed as a noninvasive and more accessible imaging modality for detecting posterior segment inflammation using machine learning techniques.
FINANCIAL DISCLOSURES: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.