TABLE 11.

Comparison with the existing machine learning-based state-of-art methods of glaucoma classification.

Author	Year	Feature extraction	Classifier	Database	No. of classes	Accuracy (%)
Li et al. (2023)	2023	ML models	SVM and RF	Private dataset	Two-class	79 for original data,
						84 for Compensated data
Khan et al. (2022)	2022	Wavelet-based	SVM	Private dataset	Two-class	91.22
		denoising and ML
Shinde. (2021)	2021	U-Net and L-Net	SVM	RIM-ONE, Drishti-GS,	Two-class	99 for L-Net,
				DRIONS-DB, JSIEC, and DRIVE		98.67 for ROI
Huang et al. (2010)	2021	Entropy-based	LDA ANN	Private dataset	Two-class	NA,
						AUC:0.95 AUC:0.97
Noronha et al. (2014)	2020	HOS cumulant	SVM and NB	Private dataset	Three-class	92.65 Average accuracy
						84.75 for mild stage
Mohamed et al. (2019)	2019	Histogram	SVM	RIM-ONE	Two-class	98.6
		and texture
Kishore and Ananthamoorthy. (2020)	2020	Intra-class and extra-class discriminative correlation analysis (IEDCA)	SVM, KNN and RF	HRF and DRIVE	Two-class	98.2 for HRF,
						97.7 for DRIVE
Proposed method	—	Pre-trained CNNs and classifier fusion	—	HVD	Three-class	85.43
				Drishti	Three-class,	90.55
				HVD + Drishti	Three-class	85.18
				ACRIMA	Two-class	99.57
				RIMONE	Two-class	94.95

Bold values are showing the model results after applying classifier fusion operation.