BMC ophthalmology · 2025
Utilization of an automated machine learning approach for the detection of granular corneal dystrophy via slit lamp photographs.
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.
Yavari Negin, Mohammadi S Saeed, Sokol Jared T, Feky Dalia El, Rajabi MohammadBagher, Hung Jia-Horung, Or Christopher, Elaraby Osama, Anover Frances A, Saengsirinavin Aim-On, Akhavanrezayat Amir, Mobasserian Azadeh, Than Ngoc Trong Tuong, Guo Jingli, Bai Yue, Yasar Cigdem, Afshari Natalie A, Lin Charles C, Nguyen Quan Dong
Аннотация
INTRODUCTION: This study aims to apply automated machine learning (AutoML) techniques for the diagnosis of granular corneal dystrophy (GCD), a rare inherited condition characterized by progressive protein deposition in the corneal stroma.
METHODS: Patients diagnosed with GCD who had slit-lamp photographs of the affected eye(s) were enrolled in the study. Individuals with concomitant corneal conditions, ungradable imaging data, or uncertain diagnoses were excluded from the study. Slit-lamp photos depicting the GCD and non-GCD were obtained from the Byers Eye Institute, Stanford University. Image processing included resizing and cropping, focusing solely on the cornea. A deep learning model was subsequently deployed, utilizing Vertex-AI, the AutoML platform developed by Google (Menlo Park, CA). The area under the precision‒recall curve (AUPRC) was plotted, and the sensitivity, specificity, positive predictive value (PPV), accuracy (AC), and F1 score were calculated.
RESULTS: The model was trained on a dataset comprising 223 images, consisting of 72 GCD and 151 non-GCD images. One hundred seventy six images were used for training, 24 were used for validation, and 23 were used for testing the model. The AUPRC for the model was 0.995 and precision and recall were both 95.70% at a confidence threshold of 0.5. The sensitivity, specificity, PPV, AC, and F1 score of the model were 93.30%, 100%, 100%, 95.70%, and 0.965, respectively.
CONCLUSIONS: A clinician-derived AutoML model successfully identified GCD from slit lamp photographs with high accuracy.
METHODS: Patients diagnosed with GCD who had slit-lamp photographs of the affected eye(s) were enrolled in the study. Individuals with concomitant corneal conditions, ungradable imaging data, or uncertain diagnoses were excluded from the study. Slit-lamp photos depicting the GCD and non-GCD were obtained from the Byers Eye Institute, Stanford University. Image processing included resizing and cropping, focusing solely on the cornea. A deep learning model was subsequently deployed, utilizing Vertex-AI, the AutoML platform developed by Google (Menlo Park, CA). The area under the precision‒recall curve (AUPRC) was plotted, and the sensitivity, specificity, positive predictive value (PPV), accuracy (AC), and F1 score were calculated.
RESULTS: The model was trained on a dataset comprising 223 images, consisting of 72 GCD and 151 non-GCD images. One hundred seventy six images were used for training, 24 were used for validation, and 23 were used for testing the model. The AUPRC for the model was 0.995 and precision and recall were both 95.70% at a confidence threshold of 0.5. The sensitivity, specificity, PPV, AC, and F1 score of the model were 93.30%, 100%, 100%, 95.70%, and 0.965, respectively.
CONCLUSIONS: A clinician-derived AutoML model successfully identified GCD from slit lamp photographs with high accuracy.