Automatic Detection of 30 Fundus Diseases Using Ultra-Widefield Fluorescein Angiography with Deep Experts Aggregation

• DOI: 10.1007/s40123-024-00900-7
• Publication Date: 2024-02-28T12:02:01Z
• Authors: Xiaoling Wang; He Li; Hongmei Zheng; Gongpeng Sun; Wenyu Wang; Zuohuizi Yi; A’min Xu; Lu He; Haiyan Wang; Wei Jia; Zhiqing Li; Chang Li; Mang Ye; Bo Du; Changzheng Chen
• Keywords: Fundus (uterus); Fundus fluorescein angiography

Inaccurate, untimely diagnoses of fundus diseases leads to vision-threatening complications and even blindness. We built a deep learning platform (DLP) for automatic detection 30 using ultra-widefield fluorescein angiography (UWFFA) with experts aggregation. This retrospective cross-sectional database study included total 61,609 UWFFA images dating from 2016 2021, involving more than 3364 subjects in multiple centers across China. All were divided into different groups. The state-of-the-art convolutional neural network architecture, ConvNeXt, was chosen as the backbone train test receiver operating characteristic curve (ROC) proposed system on data external date. compared classification performance that ophthalmologists, including two retinal specialists. DLP analyze UWFFA, which can detect up diseases, frequency-weighted average area under (AUC) 0.940 primary dataset 0.954 multi-hospital dataset. tool shows comparable accuracy retina specialists diagnosis evaluation. is first large-scale multi-retina disease classification. believe our advances by artificial intelligence (AI) various would contribute labor-saving precision medicine especially remote areas.