AUTHOR=Han Ruoan , Cheng Gangwei , Zhang Bilei , Yang Jingyuan , Yuan Mingzhen , Yang Dalu , Wu Junde , Liu Junwei , Zhao Chan , Chen Youxin , Xu Yanwu 

TITLE=Validating automated eye disease screening AI algorithm in community and in-hospital scenarios

JOURNAL=Frontiers in Public Health

VOLUME=Volume 10 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/public-health/articles/10.3389/fpubh.2022.944967

DOI=10.3389/fpubh.2022.944967

ISSN=2296-2565

ABSTRACT=Purpose: To assess the accuracy and robustness of AI algorithm for detecting referable diabetic retinopathy (RDR), referable macular diseases (RMD), and glaucoma suspect (GCS) from fundus images in community and in-hospital screening scenarios. 
Methods: We collected two color fundus image datasets, namely PUMCH (556 images, 166 subjects, four camera models) and NSDE (534 images, 134 subjects, two camera models). AI algorithm generates the screening report after taking fundus images. Three licensed ophthalmologists labeled the images as RDR, RMD, GCS, or none of the three. The resulting labels are treated as “ground truth”, and then were used to compare against the AI screening reports to validate the sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) of AI algorithm.
Results: On PUMCH dataset, regarding prediction of RDR, AI algorithm achieved overall results of 0.950±0.058, 0.963±0.024, and 0.954±0.049 on sensitivity, specificity, and AUC, respectively. For RMD, the overall results are 0.919±0.073, 0.929±0.039, and 0.974±0.009. For GCS, the overall results are 0.950±0.059, 0.946±0.016, and 0.976±0.025.
Conclusions: The AI algorithm can work robustly with various fundus camera models and achieve high accuracies for detecting RDR, RMD, and GCS.