Tan Tien-En, Ng Yi Pin, Calhoun Claire, Chaung Jia Quan, Yao Jie, Wang Yan, Zhen Liangli, Xu Xinxing, Liu Yong, Goh Rick S M, Piccoli Gabriele, Vujosevic Stela, Tan Gavin S W, Sun Jennifer K, Ting Daniel S W
Singapore Eye Research Institute, Singapore National Eye Centre, Singapore; Duke-National University of Singapore Medical School, Singapore.
Institute of High Performance Computing, Agency for Science, Technology and Research (A∗STAR), Singapore.
Ophthalmol Retina. 2025 Apr 24. doi: 10.1016/j.oret.2025.04.016.
To develop artificial intelligence (AI) models for automated detection of center-involved diabetic macular edema (CI-DME) with visual impairment using color fundus photographs (CFPs) and OCT scans.
Artificial intelligence effort using pooled data from multicenter studies.
Data sets consisted of diabetic participants with or without CI-DME, who had CFP, OCT, and best-corrected visual acuity (BCVA) obtained after manifest refraction. The development data set was from DRCR Retina Network clinical trials, external testing data set 1 was from the Singapore National Eye Centre, Singapore, and external testing data set 2 was from the Eye Clinic, IRCCS MultiMedica, Milan, Italy.
Artificial intelligence models were trained to detect CI-DME, visual impairment (BCVA 20/32 or worse), and CI-DME with visual impairment, using CFPs alone, OCTs alone, and both CFPs and OCTs together (multimodal). Data from 1007 eyes were used to train and validate the algorithms, and data from 448 eyes were used for testing.
Area under the receiver operating characteristic curve (AUC) values.
In the primary testing set, the CFP model, OCT model, and multimodal model had AUCs of 0.848 (95% confidence interval [CI], 0.787-0.900), 0.913 (95% CI, 0.870-0.947), and 0.939 (95% CI, 0.906-0.964), respectively, for detection of CI-DME with visual impairment. In external testing data set 1, the CFP, OCT, and multimodal models had AUCs of 0.756 (95% CI, 0.624-0.870), 0.949 (95% CI, 0.889-0.989), and 0.917 (95% CI, 0.837-0.979), respectively, for detection of CI-DME with visual impairment. In external testing data set 2, the CFP, OCT, and multimodal models had AUCs of 0.881 (95% CI, 0.822-0.940), 0.828 (95% CI, 0.749-0.905), and 0.907 (95% CI, 0.852-0.952), respectively, for detection of CI-DME with visual impairment.
The AI models showed good diagnostic performance for the detection of CI-DME with visual impairment. The multimodal (CFP and OCT) model did not offer additional benefit over the OCT model alone. If validated in prospective studies, these AI models could potentially help to improve the triage and detection of patients who require prompt treatment.
FINANCIAL DISCLOSURE(S): Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.
利用彩色眼底照片（CFP）和光学相干断层扫描（OCT）扫描，开发用于自动检测伴有视力损害的中心性糖尿病性黄斑水肿（CI-DME）的人工智能（AI）模型。
使用来自多中心研究的汇总数据进行人工智能研究。
数据集包括患有或未患有CI-DME的糖尿病参与者，他们有CFP、OCT以及在显验光后获得的最佳矫正视力（BCVA）。开发数据集来自糖尿病视网膜病变临床研究网络（DRCR）视网膜网络临床试验，外部测试数据集1来自新加坡国立眼科中心，外部测试数据集2来自意大利米兰IRCCS MultiMedica眼科诊所。
训练人工智能模型，分别使用单独的CFP、单独的OCT以及CFP和OCT两者（多模态）来检测CI-DME、视力损害（BCVA为20/32或更差）以及伴有视力损害的CI-DME。来自1007只眼睛的数据用于训练和验证算法，来自448只眼睛的数据用于测试。
受试者操作特征曲线（ROC）下面积（AUC）值。
在主要测试集中，对于检测伴有视力损害的CI-DME，CFP模型、OCT模型和多模态模型的AUC分别为0.848（95%置信区间[CI]，0.787-0.900）、0.913（95%CI，0.870-0.9"
47）和0.939（95%CI，0.906-0.964）。在外部测试数据集1中，对于检测伴有视力损害的CI-DME，CFP、OCT和多模态模型的AUC分别为0.756（95%CI，0.624-0.870）、0.949（95%CI，0.889-0.989）和0.917（95%CI，0.837-0.979）。在外部测试数据集2中，对于检测伴有视力损害的CI-DME，CFP、OCT和多模态模型的AUC分别为0.881（95%CI，0.822-0.940）、0.828（95%CI，0.749-0.905）和0.907（95%CI，0.852-0.952）。
AI模型在检测伴有视力损害的CI-DME方面显示出良好的诊断性能。多模态（CFP和OCT）模型相比单独的OCT模型没有额外优势。如果在前瞻性研究中得到验证，这些AI模型可能有助于改善对需要及时治疗的患者的分诊和检测。
在本文末尾的脚注和披露中可能会发现专有或商业披露信息。
