Table 8.

Comparison of different fusion strategies on external test datasets.

External dataset		FN-F1-OCT		FN-Weight-OCT		FN-Auto-OCT
		Complete model	Limited model	Complete model	Limited model	Complete model	Limited model
ACC	CNV	0.7833	0.9167	0.8167	0.9167	0.8167	0.95
	DME	0.785	0.7663	0.757	0.7477	0.7851	0.9065
	Drusen	0.5185	0.5556	0.4445	0.2593	0.4074	0.7407
	Normal	0.7349	1	0.6747	0.9639	0.9639	0.9639
	Weighted avg	0.74	0.85	0.71	0.8	0.81	0.92
Recall	CNV	0.7833	0.9167	0.8167	0.9167	0.8167	0.95
	DME	0.785	0.7663	0.757	0.7477	0.7851	0.9065
	Drusen	0.5185	0.5556	0.4445	0.2593	0.4074	0.7407
	Normal	0.7349	1	0.6747	0.9639	0.9639	0.9639
	Weighted avg	0.74	0.85	0.71	0.8	0.81	0.92
F1	CNV	0.7768	0.8397	0.7968	0.7746	0.7778	0.9120
	DME	0.7636	0.8677	0.7232	0.8377	0.8442	0.9327
	Drusen	0.5714	0.6250	0.5455	0.3415	0.4889	0.7145
	Normal	0.7439	0.8925	0.6871	0.8889	0.8695	0.9697
	Weighted avg	0.74	0.85	0.71	0.79	0.8	0.92
AUC	CNV	0.859	0.921	0.876	0.896	0.869	0.957
	DME	0.807	0.883	0.773	0.862	0.869	0.942
	Drusen	0.743	0.766	0.712	0.616	0.69	0.852
	Normal	0.816	0.948	0.775	0.938	0.928	0.977
	Weighted avg	0.829	0.899	0.81	0.868	0.872	0.945