Table 5.

Comparison of the proposed method with the state-of-the-art.

Paper	Classifier	Best Score (Accuracy)	XAI Method	Dataset
[15]	Support Vector Machines, KNN, MLP	91.4%	LIME, SHAP	Dementia dataset
[16]	CNN	95.4%	HAM, PCR	MRI scans ADNI
[17]	Graph Neural Network (GNN)	53.5 ± 4.5%	GNN Explainer	ADNI
[18]	EfficientNetB0	80%	Occlusion Sensitivity Mapping	MRI scans OASIS
[19]	3D CNN	-	Saliency Map, LRP	18F-FDG PET
[20]	KNN, RF, AdaBoost, Gradient Boosting Bernouli NB, SVM	91%	DT	Cognitive and and PET images
[21]	3D CNN	76.6%	3D Ultrametric Contour Map, 3D Class Activation Map, 3D GradCAM	ADNI
[22]	3D CNN	77%	Sensitivity Analysis Occlusion	MRI, PET
Proposed Method	VGG16, VGG19, DenseNet169, DenseNet201, Ensemble 1 (VGG16, VGG19) Ensemble 2 (DenseNet169, DenseNet20), Proposed model (EfficientNetB3 & CNN)	96%	Saliency maps, Grad-CAM (Gradient-weighted Class Activation Mapping)	MRI scans OASIS