Retinal photograph-based deep learning system for detection of hyperthyroidism: a multicenter, diagnostic study

Dong, Li; Ju, Lie; Hui, Shiqi; Luo, Lihua; Jiang, Xue; Nie, Zihan; Zhang, Ruiheng; Zhou, Wenda; Li, Heyan; Jonas, Jost B.; Wang, Xin; Zhao, Xin; He, Chao; Chen, Yuzhong; Wang, Zhaohui; Gao, Jianxiong; Ge, Zongyuan; Wei, Wenbin; Li, Dongmei

doi:10.1186/s40537-023-00777-6

Journal of Big Data

Table 2 Performance of the DL-based system for the prediction of hyperthyroidism from retinal photographs

From: Retinal photograph-based deep learning system for detection of hyperthyroidism: a multicenter, diagnostic study

	Internal validation dataset		BTH	BFH	CZMC	SYMC	SZMC	WJMC
	Model 1	Model 2	External validation dataset 1^a	External validation dataset 2^a	External validation dataset 3^a	External validation dataset 4^a	External validation dataset 5^a	External validation dataset 6^a
AUC (95% CI)	0.907 (0.894–0.918)	0.850 (0.832–0.866)	0.816 (0.789–0.846)	0.823 (0.787–0.858)	0.838 (0.810–0.864)	0.822 (0.796–0.847)	0.849 (0.824–0.874)	0.816 (0.791–0.844)
Accuracy (95% CI)	0.809 (0.790–0.826)	0.780 (0.758–0.801)	0.735 (0.706–0.767)	0.735 (0.700–0.773)	0.796 (0.765–0.824)	0.762 (0.729–0.792)	0.789 (0.757–0.818)	0.755 (0.722–0.785)
Sensitivity (95% CI)	0.761 (0.721–0.796)	0.718 (0.674–0.758)	0.625 (0.567–0.680)	0.617 (0.543–0.687)	0.681 (0.613–0.741)	0.674 (0.609–0.733)	0.738 (0.677–0.792)	0.671 (0.606–0.730)
Specificity (95% CI)	0.885 (0.862–0.904)	0.824 (0.799–0.847)	0.845 (0.797–0.883)	0.852 (0.791–0.899)	0.833 (0.797–0.864)	0.799 (0.761–0.833)	0.814 (0.776–0.847)	0.795 (0.756–0.829)
Positive predictive value (95% CI)	0.791 (0.752–0.825)	0.659 (0.616–0.700)	0.801 (0.742–0.849)	0.807 (0.730–0.867)	0.634 (0.568–0.695)	0.613 (0.550–0.673)	0.658 (0.597–0.714)	0.609 (0.546–0.668)
Negative predictive value (95% CI)	0.866 (0.842–0.887)	0.861 (0.836–0.882)	0.693 (0.642–0.739)	0.690 (0.625–0.749)	0.860 (0.826–0.889)	0.838 (0.801–0.870)	0.865 (0.830–0.894)	0.835 (0.798–0.867)
F1 score	0.837 (0.802–0.870)	0.813 (0.797–0.831)	0.736 (0.706–0.767)	0.735 (0.700–0.773)	0.788 (0.763–0.813)	0.759 (0.731–0.784)	0.789 (0.763–0.814)	0.755 (0.730–0.781)

Data are presented as estimate (95% Confidence interval)
AUC area under the curve, CI confidence interval, BTH beijing tongren hospital, BFH Beijing friendship hospital, CZMC: chongqing zhuoyue medical centre of ikang healthcare group, SYMC Shanghai Yuanhua Medical Centre of iKang Guobin Healthcare Group, SZMC: Shenzhen Zhuoyue medical centre of ikang healthcare group, WJMC Wuhan Jindun Road Medical Centre of iKang Healthcare Group
^aOnly model 1 was applied in external validation

Back to article page