TABLE III.

Comparison With Existing Studies Using MRI Data of ADNI for AD vs. NC Classification

Method	Feature	Classifier	Subjects	Template	ACC (%)	SEN (%)	SPE (%)
Cuingnet et al. [15]	Voxel-direct-D GM	SVM	137 AD + 162 NC	Single	88.58	81.00	95.00
Zhang et al. [16]	93 ROI GM	SVM	51 AD + 52 NC	Single	86.20	86.00	86.30
Zhang et al. [14]	93 ROI GM	SVM	91 MCI + 50 NC	Single	84.80	–	–
Liu et al. [50]	Voxel-wise GM	SRC ensemble	198 AD + 229 NC	Single	90.80	86.32	94.76
Liu et al. [13]	Voxel-wise GM	SVM ensemble	198 AD + 229 NC	Single	92.00	91.00	93.00
Eskildsen et al. [58]	ROI-wise cortical thickness	LDA	194AD + 226NC	Single	84.50	79.40	88.90
Cho et al. [59]	Cortical thickness	PCA-LDA	128 AD + 160 NC	Single	–	82.00	93.00
Coupé et al. [60]	Hippocampus and entorhinal cortex volume and grading	QDA	60 AD + 60 NC	Single	90.00	88.00	92.00
Duchesne et al. [10]	Tensor-based morphometry	SVM	75 AD + 75 NC	Single	92.00	–	–
Koikkalainen et al. [18]	Tensor-based morphometry	Linear regression	88AD + 115NC	Multiple	86.00	81.00	91.00
Wolz et al. [30]	Four MR features	LDA	198 AD + 231 NC	Multiple	89.00	93.00	85.00
Min et al. [17]	Data-driven ROI GM	SVM	97AD + 128 NC	Multiple	91.64	88.56	93.85
Min et al. [29]	Data-driven ROI GM	SVM	97AD + 128NC	Multiple	90.69	87.56	93.01
Liu et al. [28]	Data-driven ROI GM	SVM ensemble	97AD + 128NC	Multiple	92.51	92.89	88.33
Proposed	Data-driven ROI GM	SVM ensemble	97AD + 128NC	Multiple	93.06	94.85	90.49

Note: SVM means Support Vector Machine; SRC denotes Sparse Regression Classifier; LDA represents Linear Discriminant Analysis; PCA-LDA denotes Principal Component Analysis-Linear Discriminant Analysis; QDA denotes Quadratic Discriminant Analysis.