A Novel Context Aware Joint Segmentation and Classification Framework for Glaucoma Detection

Table 5

Classification results comparison with state-of-the-art for four test datasets.

Dataset	Model	Accuracy	Sensitivity	Specificity	AUC
ACRIMA	GD-YNet (proposed)	0.9972	1.0	0.9935	1.0
	XceptionNet [11]; Diaz-Pinto et al. (2019)	0.7021	—	—	0.7678
	DenseNet [43]; Sreng et al. (2020)	0.9953	—	—	0.9998
	AlexNet [52]; Taj et al. (2021)	0.995	1.0	0.989	—
	InceptionV3 [52]; Taj et al. (2021)	0.985	0.991	0.977	—
	InceptionResNetV2 [52]; Taj et al.(2 021)	0.990	1.0	0.977	—
	NasNet-Large [52]; Taj et al. (2021)	0.995	0.991	1.0	—
Drishti-gs	GD-YNet (proposed)	0.9802	1.0	0.9355	1.0
	OverFeat and VGG-S [12]; Orlando et al. (2017)	—	—	—	0.7626
	XceptionNet [11]; Diaz-Pinto et al. (2019)	0.7525	—	—	0.8041
	ShuffleNet [43]; Sreng et al. (2020)	0.8667	—	—	0.7884
REFUGE	GD-YNet proposed	0.9950	1.0	0.9944	1.0
	Multitask CNN [25]; Chakravartyand Sivswamy (2018)	—	0.88	0.91	0.96
	Ensemble +SVM [43]; Sreng et al. (2020)	0.9575	—	—	0.9432
RIMOne-V1	GD-YNet proposed	0.9941	1.0	0.9915	1.0
	AlexNet [52]; Taj et al. (2021)	0.875	0.836	0.904	—
	InceptionV3 [52]; Taj et al. (2021)	0.922	0.891	0.945	—
	InceptionResNetV2 [52]; Taj et al. (2021)	0.906	0.855	0.945	—
	NasNet-Large [52]; Taj et al. (2021)	0.945	0.927	0.959	—
	VGG19 [14]; Gómez-Valverde et al. (2019)	—	0.8701	0.8901	0.94
	XceptionNet [11]; Diaz-Pinto et al. (2019)	0.7121	—	—	0.8575
	SqueezeNet [43]; Sreng et al. (2020)	0.9737	—	—	1.0

-Not reported.