Table 2.

Some recent works on OCT image classification considering diabetic retinopathy (MCME = multi-scale convolutional mixture of experts, LBP = local binary patterns, BoF = bag of features, SURF = speeded up robust features, and MLP = multilayer perceptron).

Author	Dataset	Image type	Method	Results
Lemaître et al.24	SERI Dataset of normal, DME-cyst, DME-Exudate	SD-OCT images	LBP features, different mapping strategies and using linear and nonlinear classifiers	Sensitivity = 81.2% Specificity = 93.7%
Rasti et al.25	Duke university dataset and private dataset of AMD, DME, and normal	Macular OCT	MCME model containing CNN experts and gating network are fed by specific scales of the input pattern	Precision = 98.86%
				AUC-ROC = 99.85%
Shih et al.26	Dataset of DME, CNV, Drusen, and normal30	OCT images	Pre-trained VGG16	Accuracy = 99.48%
Tsuji et al.27	Dataset of DME, CNV, Drusen, and normal30	OCT images	Capsule network	Accuracy = 99.6%
Kazeminasab et al.29	Dataset of MA and normal OCT strips	OCT images	BoF and SURF, MLP classifier	Accuracy = 94.5%