Table 2.

A comparison of some current studies with respect to accuracy (ACC), specificity (SPE), sensitivity (SEN), Dice similarity coefficient (DSC), Quadratic Kappa Score (QKS) and the area under the curve (AUC)

Study	Year	Analysis	Methods	Dataset	Aug.	Performance
Maninis et al. [18]	2016	BV and OD segmentation	CNN	DRIVE, STARE for BV segmentation, DRIONS-DB, RIM-ONE (r3) for OD	No	DRIVE DSC 82.2%, STARE DSC=83.1%, DRIONS-DB DSC 97.1%, RIM-ONE (r3) DSC 95.9%
Islam et al. [19]	2018	MA detection, DR grading based MA	CNN	KAGGLE	Yes	QKS 85.1%, AUC 84.4%,SEN 98%, SPE 94%
Eftekhari et al. [20]	2019	MA segmentation	CNN	E-Ophtha-MA	Yes	SEN 80%
Gurani et al. [21]	2019	DR detection	CNN	KAGGLE	No	recognition rate 93%
Khalifa et al. [22]	2019	DR grading	DenseNet	APTOS 2019	Yes	ACC 97.7%
Hagos and Kant [23]	2019	DR detection	Inception-V3	KAGGLE	Yes	ACC 90.9%
Abdelmaksoud et al. [24]	2020	EX, MA, HM, BV segmentation and DR grading	matched filter with first order gaussian derivative, morphological operation and MLSVM	DRIVE, STARE, MESSIDOR, IDRiD	No	ACC 89.2%, AUC 85.20%, SEN 85.1%, SPE 85.2%,PPV 92.8%, DSC 88.7%
Patil et al. [25]	2020	DR grading	CNN	Kaggle, MESSIDOR	No	ACC 89.1%
Nazir et al. [26]	2020	OD, CD, HM, EX, and MA segmentation	FRCNN, and FKM	ORIGA, MESSIDOR, HRF, DiaretDB1	No	mAP 94%
Shah et al. [27]	2020	DR Detection	CNN	MESSIDOR	No	SEN 99.7%, SPE 98.5%, AUC 99.1%, QKS 0.95
Tymchenko et al. [28]	2020	DR Detection	ensemble of (EfficientNet, SE-ResNeXt50)	APTOS 2019	Yes	SEN 99%, SPE 99%, QKS 0.92
Abdelmaksoud et al. [29]	2021	EX, MA, HM, BV segmentation and DR grading	U-Net and MLSVM	DRIVE, STARE, MESSIDOR, IDRiD, ChaseDB1, HRF, DIARETDB1, E-ophtha, DIARETDB0	No	ACC 95.1%, AUC 91.9%, SEN 86.1%, SPE 86.8%, PPV 84.7%, DSC 86.2%
Aswathi et al.[30]	2021	DR grading	Inception-V3	MESSIDOR	Yes	ACC 78% (0,1), 69% (0,2), 61% (1,2), 62% (1,3), 49% (2,3), 32% (0,3)