. 2021 Apr 29;35(9):2456–2457. doi: 10.1038/s41433-021-01561-7

Table 1.

Openly accessible fundus photography datasets using in machine learning models for glaucoma diagnosis.

Dataset	Fundus photographs		Confirmation by visual field	Confirmation by clinical exam	Presence of myopes	Presence of physiological cupping	Presence of normal tension glaucoma	Total sample size	Glaucoma sample size	Severity assessment
Dataset	Disc	Maula	Confirmation by visual field	Confirmation by clinical exam	Presence of myopes	Presence of physiological cupping	Presence of normal tension glaucoma	Total sample size	Glaucoma sample size	Severity assessment
ORIGA	Yes	No	No	No	No	No	No	650	168	No
RIGA	Yes	Yes	No	No	No	No	No	750	0	No
REFUGE	Yes	No	Yes	Yes	Yes	No	Yes	1200	120	No
ACRIMA	Yes	No	No	Yes	No	No	No	705	396	No
RIM-ONEv1	Yes	No	No	Yes	No	No	Yes	169	51	Yes
RIM-ONEv2	Yes	No	No	Yes	No	No	No	452	197	No
RIM-ONEv3	Yes	No	No	Yes	No	No	No	127	35	No
DRISHTI-GS1	Yes	No	No	Yes	No	No	No	101	80	No
High-Resolution Fundus (HRF)	Yes	Yes	No	Yes	No	No	No	45	15	No
LES-AV	Yes	No	No	Yes	No	No	Yes	22	11	No