Table 7.

Summary of the ML-related studies in AMD diagnosis.

Study	Method	Dataset	Classification	Performance	Modality
Fraccaro et al. [114]	White box (decision trees, logistic regression), Black box (AdaBoost, random forests, SVM)	EHR data from Genoa, Italy	AMD diagnosis	AUC: 0.90–0.92	OCT
Mookiah et al. [12]	Feature extraction (PO, CC), Classification (DT, k-NN, NB, PNN, SVM)	ARIA, STARE datasets	Normal vs. AMD	Accuracy: 91.36–97.78%	CFP
Phan et al. [111]	Color, texture, context analysis, Random forest, SVM	Database of 279 fundus photographs	AMD stage classification	AUC: 0.739–0.874	CFP
Alfahaid et al. [40]	KNN classifier using rotation-invariant uniform local binary pattern texture characteristics	Manchester Royal Eye Hospital	AMD and healthy	Accuracy: 89% (all layers) 89% (superficial) 94% (deep) 98% (outer) 100% (choriocapillaris)	OCTA
Wang et al. [10]	Feature extraction (LCP), Classification (BP, SMO, SVM, LR, NBayes, RF)	OCT dataset from multiple universities	AMD vs. DME vs. healthy macula	Accuracy: 99.3%	OCT
Nugroho et al. [115]	Feature extraction (DenseNet, ResNet50, LBP, HOG), Classification (Logistic regression)	OCT images	Normal vs. DME vs. Drusen vs. CNV	Accuracy: 88–89%	OCT
Hussain et al. [11]	Feature extraction (retinal parameters), Classification (Random Forest)	SD-OCT images	DME vs. AMD	Accuracy: >95%	OCT
Li et al. [23]	Feature integration (RC Net), Classification (RC Net)	OCT images	Eye disease categorization	Accuracy: 99.6%	OCT
Govindaiah et al. [110]	ML and statistical algorithms (Random Forest, Naïve Bayes, Logistic model tree, etc.)	AREDS study data	Late AMD prediction	Accuracy: 72.9%, Sensitivity: 73.8%, Specificity: 72.7%	CFP