An Ensemble-of-Classifiers Based Approach for Early Diagnosis of Alzheimer’s Disease: Classification Using Structural Features of Brain Images

Table 3

Comparison of proposed approach with existing approaches.

Approach	Features	Classifier	Accuracy		Specificity	Sensitivity
Ye et al., 2008 [27]	ROI and voxel based tensor	RKDA	—		89.50	95.00
		SVM	—		85.00	94.50
Long and Wyatt, 2010 [28]	WM	Quick shift clustering	94.67	96		—
	GM		97.33		—
Kloppel et al., 2008 [6]	GM	SVM	95.6		94.1	97.1
Zhang et al., 2011 [29]	GM volume (93 ROIs)	SVM	86.2		86.3	86
Casanova et al., 2013 [30]	GM-voxel	RLR	87.1		88.9	84.3
Chu et al., 2012 [31]	GM-voxel	SVM	84.3		—	—
Cuingnet et al., 2011 [12]	GM-voxel	SVM	88.6		95	81
Vemuri et al., 2008 [32]	GM + WM + CSF voxels	SVM	—		86.0	86.0
Wee et al., 2013 [19]	Correlation and ROI based morphological features	SVM	92.35		94.31	90.35
Teipel et al., 2007 [33]	GM + WM	Logistic regression	83		78	88
Westman et al., 2013 [34]	Regional MRI measures (259 features)	OPLS	91.5		92.9	89.8
Hinrichs et al., 2009 [14]	GM-voxels	LP boosting	82		80	85
Wolz et al., 2011 [13]	Hippocampus volume, tensor-based morphometry, cortical thickness, manifold learning based features	LDA	89		93	85
Liu et al., 2014 [35]	GM-voxel	Hierarchical fusion	92		93	90.9
Proposed approach	ROI (left hippocampus)	Ensemble of classifiers	93.75	90.5	100	87.5
	Volume of GM		87.5		100	75