Skip to main content

Table 3 Studies selected following PRISMA guidelines presented in chronological order

From: Machine learning methods for predicting progression from mild cognitive impairment to Alzheimer’s disease dementia: a systematic review

Author (year) Groups Sample size (mean age) Database Neuroimaging technique and features Classification method Validation method Results (% accuracy) AUC ROC
Plant et al. (2010) [24] HS
AD
MCI
18 (64.8)
32 (68.8)
24 (69.7)
Sample collected for the study MRI: whole-brain volume measures SVM
Bayes
VFI
Train/test method: AD + HS as train set and MCI as test set. pMCI vs sMCI:
SVM: 50
Bayes: 58.3
VFI: 75
NA
Chincarini et al. (2011) [25] HS
AD
sMCI
pMCI
189 (76.6)
144 (75.5)
166 (75.7)
136 (75.1)
ADNI-1 MRI: GM volumes SVM 20-fold Cross Validation NA 0.74
Costafreda et al. (2011) [26] HS
AD
MCI
88 (73.6)
71 (74.9)
103 (74.1)
AddNeuroMed MRI: 3D hippocampal morphometric measures SVM with RBF kernel 4-fold Cross Validation pMCI vs sMCI: 80 NA
Filipovych et al. (2011) [27] HS
AD
sMCI
pMCI
63 (75.2)
54 (77.4)
174 (74.5)
68 (76.2)
ADNI-1 MRI: whole-brain GM density Semi-supervised SVM Leave-one-out Cross Validation pMCI: 79.4
sMCI: 51.7
0.69
Hinrichs et al. (2011) [8] HS
AD
MCI
66 (76.2)
48 (76.6)
119 (75.1)
ADNI-1 MRI and PET: scan data, APOE4 genotype, CSF assays, and cognitive tests MK-SVM Train/test method: AD + HS as train set and MCI as test set pMCI vs sMCI: NA 0.79
Westman et al. (2011) [28] HS
AD
MCI
112 (73)
117 (76)
122 (75)
AddNeuroMed MRI: whole-brain volume, age, and education OPLS Train/test method: sample of 51 subjects pMCI vs sMCI: 73 NA
Wolz et al. (2011) [29] HS
AD
sMCI
pMCI
231 (76)
198 (75.7)
238 (74.8)
167 (74.6)
ADNI-1 MRI: hippocampal volume, cortical thickness, tensor-based morphometry, and manifold-based learning SVM
LDA
Train/test method: 95/5 partition pMCI vs sMCI:
SVM: 60
LDA: 68
NA
Zhang et al. (2011) [30] HS
AD
sMCI
pMCI
52 (75.3)
51 (75.2)
56 (75.3)
43 (75.3)
ADNI-1 MRI and PET: volume, intensity, and CSF (Aβ42, t-tau y p-tau) measures SVM 10-fold cross-validation pMCI: 91.5
sMCI: 73.4
NA
Batmanghelich et al. (2012) [31] sMCI
pMCI
139 (NA)
99 (NA)
ADNI-1 MRI: WM, GM, and CSF Logistic model trees + Laplacian SVM 5-fold cross-validation pMCI vs sMCI: 61.5 NA
Cheng et al. (2012) [32] HS
AD
sMCI
pMCI
52 (75.3)
51 (75.2)
56 (75.3)
43 (75.3)
ADNI-1 MRI and PET: GM and WM volume, intensity, and CSF (Aβ42, t-tau y p-tau) measures Domain Transfer SVM Train/test method: AD + HS as train set and MCI as test set with 10-fold cross-validation pMCI vs sMCI: 69.4 0.74
Cho et al. (2012) [33] HS
AD
sMCI
pMCI
160 (76.2)
128 (76)
131 (74.1)
72 (74.8)
ADNI-1 MRI: cortical thickness LDA Train/test method: 50/50 partition pMCI vs sMCI: 70 NA
Gray et al. (2012) [34] HS
AD
sMCI
pMCI
54 (NA)
50 (NA)
64 (NA)
53 (NA)
ADNI-1 PET: signal intensity and relative change over 12 months SVM with RBF kernel Train/test method: 75/25 partition with 1000 iterations pMCI vs sMCI: 63.1 0.66
Li et al. (2012) [35] HS
AD
sMCI
pMCI
40 (73.7)
37 (74.8)
36 (75.3)
39 (75.6)
ADNI-1 MRI: static and dynamic cortical thickness and clustering coefficient SVM Leave-one-out cross-validation pMCI vs sMCI: 81.7 NA
Toussaint et al. (2012) [36] HS
AD
sMCI
pMCI
80 (76.4)
80 (76)
40 (76.4)
40 (76.4)
ADNI-1 PET: glucose metabolic signal and clinical measures Two-sample t-test + spatial ICA + SVM with RBF kernel Leave-one-out cross-validation pMCI vs sMCI: 80 NA
Wee et al. (2012) [37] HS
MCI
17 (72.1)
10 (74.2)
ADNI-1 MRI and PET: WM structural connectivity and GM functional connectivity Mk-SVM Leave-one-out cross-validation pMCI vs sMCI: 96.3 0.95
Ye et al. (2012) [38] sMCI
pMCI
177 (NA)
142 (NA)
ADNI-1 MRI: GM and WM volumes, cortical thickness, demographic, genetic, and cognitive measures SVM Leave-one-out Cross Validation pMCI vs sMCI: NA 0.85
Zhang et al. (2012) [9] HS
AD
sMCI
pMCI
50 (75.3)
45 (75,4)
48 (74.7)
43 (75.8)
ADNI-1 MRI and PET: volume, intensity, and CSF (Aβ42, t-tau y p-tau) measures M3TL 10-fold cross-validation pMCI vs sMCI: 73.9 0.80
Adaszewski et al. (2013) [39] HS
AD
sMCI
pMCI
137 (NA)
108 (NA)
61 (74)
142 (74)
ADNI-1
ADNI-GO
MRI: whole-brain GM volume SVM Train/test method: AD + HS subset as train set and bootstrapping with 100 permutations pMCI: 63.7
sMCI: NA
NA
Aguilar et al. (2013) [40] HS
AD
sMCI
pMCI
110 (73)
116 (74.4)
98 (74.7)
21 (72.9)
AddNeuroMed MRI: volume and cortical thickness OPLS
SVM
Decision Trees
ANN
10-fold cross-validation pMCI vs sMCI:
OPLS: 74.7
SVM: 70.9
Decision Trees: 67.4
ANN: 70.1
0.83
0.81
0.80
0.75
Babu et al. (2013) [41] HS
sMCI
pMCI
232 (76)
236 (74.9)
167 (74.6)
ADNI-1 MRI: GM volumes PBL-McqRBFN Train/test method: 95/5 partition pMCI vs sMCI: 79 NA
Casanova et al. (2013) [42] HS
AD
sMCI
pMCI
188 (75.9)
171 (75.5)
182 (75.2)
153 (75)
ADNI-1 MRI: GM volumes RLR 10-fold cross-validation pMCI vs sMCI: 61.5 NA
Cheng et al. (2013) [43] HS
AD
sMCI
pMCI
52 (NA)
51 (NA)
56 (NA)
43 (NA)
ADNI-1 MRI and PET: volume, intensity, APOE4 genotype, and CSF (Aβ42, t-tau y p-tau) measures SM2TLC 10-fold cross-validation pMCI vs sMCI: 77.8 0.81
Liu, M. et al. (2013) [44] sMCI
pMCI
185 (74.9)
164 (74.9)
ADNI-1 MRI: GM volumes MTSRC Leave-one-out cross-validation pMCI vs sMCI: 74.1 0.75
Liu, X. et al. (2013) [45] HS
AD
sMCI
pMCI
138 (76)
86 (75)
93 (75)
97 (75)
ADNI-1 MRI: volume and cortical thickness SVM
EN
LDA
Leave-one-out cross-validation pMCI vs sMCI:
SVM: 66
EN: 68
LDA: 68
0.53
NA
0.68
Wee et al. (2013) [46] HS
AD
sMCI
pMCI
200 (75.8)
198 (75.7)
111 (75.3)
89 (74.8)
ADNI-1 MRI: cortical thickness and correlation of cortical thickness between pairs of ROIs Mk-SVM 10-fold cross-validation pMCI vs sMCI: 75 0.84
Young et al. (2013) [47] HS
AD
sMCI
pMCI
73 (75.9)
63 (75.2)
96 (75.6)
47 (74.5)
ADNI-1 MRI and PET: volume, intensity, APOE4 genotype, and CSF (Aβ42, t-tau y p-tau) measures Gaussian Process Leave-one-out cross-validation pMCI vs sMCI: 74.1 0.80
Apostolova et al. (2014) [48] HS
AD
MCI
111 (NA)
95 (NA)
182 (NA)
ADNI-1 MRI: hippocampal volumes and demographic, APOE genotype, and CSF measures SVM Leave-one-out cross-validation pMCI vs sMCI: 68 0.68
Guerrero et al. (2014) [49] HS
AD
sMCI
pMCI
175 (76.3)
106 (75.4)
114 (75.1)
116 (74.7)
ADNI-1
ADNI-GO
MRI: 3D brain volumes SVM Train/test method: unknown partition pMCI vs sMCI: 97.2 NA
Lebedev et al. (2014) [50] HS
AD
MCI
225 (75.9)
185 (75.2)
165 (75.5)
ADNI-1 MRI: cortical thickness, demographic variables, and APOE4 genotype RF Independent test set pMCI vs sMCI: 82.3 0.83
Liu, M. et al. (2014) [51] HS
AD
sMCI
pMCI
229 (76)
198 (75.7)
236 (74.9)
167 (74.9)
ADNI-1 MRI. whole-brain GM density SVM 10-fold cross-validation pMCI vs sMCI: 70.7 NA
Liu, F. et al. (2014) [52] HS
AD
MCI
52 (75.3)
51 (75.2)
99 (75.3)
ADNI-1 MRI and PET: volume and intensity measures Mk-SVM 10-fold cross-validation pMCI vs sMCI: 67.8 0.70
Min et al. (2014) [53] HS
AD
sMCI
pMCI
128 (76.1)
97 (75.9)
117 (75.1)
117 (75.2)
ADNI-1 MRI: multi-atlas GM volume measures SVM 10-fold cross-validation pMCI vs sMCI: 72.4 0.67
Suk et al. (2014) [54] HS
AD
MCI
101 (75.9)
93 (75.5)
204 (74.9)
ADNI-1 MRI and PET: volume and intensity measures DBM 10-fold cross-validation pMCI vs sMCI: 75.9 0.75
Tong et al. (2014) [55] HS
AD
sMCI
pMCI
231 (76)
198 (75.7)
238 (74.9)
167 (74.6)
ADNI-1 MRI: intensity patches Multiple instance-graph Leave-one-out cross-validation pMCI vs sMCI: 70.4 NA
Cabral et al. (2015) [56] sMCI
pMCI
56 (NA)
44 (NA)
ADNI-1 PET: voxel intensities Linear-SVM
SVM-RBF
Gaussian Naïve Bayes
10-fold cross-validation pMCI vs sMCI:
Linear-SVM: 71–89
SVM-RBF: 75–85
Gaussian Naïve Bayes: 73–81
NA
Cheng et al. (2015) [57] HS
AD
sMCI
pMCI
52 (NA)
51 (NA)
56 (NA)
53 (NA)
ADNI-1 MRI and PET: volume, intensity and CSF (Aβ42, t-tau y p-tau) measures M2TL 10-fold cross-validation pMCI vs sMCI: 80.1 0.85
Cheng et al. (2015) [58] HS
AD
sMCI
pMCI
52 (NA)
51 (NA)
56 (NA)
43 (NA)
ADNI-1 MRI and PET: volume, intensity, APOE4 genotype, and CSF (Aβ42, t-tau y p-tau) measures Domain Transfer SVM 10-fold cross-validation pMCI vs sMCI: 79.4 0.85
Moradi et al. (2015) [59] HS
AD
sMCI
pMCI
231 (NA)
200 (NA)
100 (NA)
164 (NA)
ADNI-1 MRI: GM volumes, age, and cognitive measures RF 10-fold cross-validation pMCI vs sMCI: 81 0.90
Raamana et al. (2015) [60] HS
AD
sMCI
pMCI
159 (NA)
136 (NA)
130 (NA)
56 (NA)
ADNI-1 MRI: cortical thickness Variational Bayes probabilistic MKL Train/test method: 95/5 partition pMCI vs sMCI: 64 0.68
Ritter et al. (2015) [61] sMCI
pMCI
151 (74.1)
86 (74.6)
ADNI-1 MRI and PET: cortical thickness, intensity measurements, neuropsychological tests, clinical variables, and demographic data SVM with RBF kernel
Classification tree
RF
30 iterations of 10-fold cross-validation pMCI vs sMCI: SVM: 61-64
Classification Tree: 61-64
RF: 60-61
NA
Salvatore et al. (2015) [62] HS
AD
sMCI
pMCI
162 (76.3)
137 (76)
134 (74.5)
76 (74.8)
ADNI-1 MRI: GM and WM volumes SVM 20-fold cross-validation pMCI vs sMCI: 66 NA
Xu et al. (2015) [63] HS
AD
MCI
117 (75.4)
113 (75.6)
110 (75.2)
ADNI-1 MRI and PET: volume and intensity measures wmSRC 10-fold cross-validation pMCI vs sMCI: 77.8 0.80
Ardekani et al. (2016) [64] sMCI
pMCI
78 (NA)
86 (NA)
ADNI-1 MRI: hippocampal volumetric integrity, APOE genotype, demographic data, and neuropsychological tests RF Out-of-bag method pMCI vs sMCI: 82.3 0.83
Collij et al. (2016) [66] HS
AD
MCI
100 (61.7)
100 (63.1)
60 (62.7)
Sample collected for the study MRI: whole-brain arterial spin labeling perfusion maps SVM Train/test method: 50/50 partition pMCI vs sMCI: 70.8 0.77
Li et al. (2016) [67] HS
AD
sMCI
pMCI
42 (65.6)
25 (69.4)
10 (66.5)
21 (68.6)
ADNI-1 MRI: GM whole-brain and seed-based functional connectivity SVM Leave one out cross-validation pMCI vs SMCI: 80.6 NA
Liu et al. (2016) [68] HS
AD
sMCI
pMCI
97 (75.9)
128 (76.1)
117 (75.1)
117 (75.2)
ADNI-1 MRI: GM density maps SVM 10-fold cross-validation pMCI vs SMCI: 79.2 0.83
López et al. (2016) [69] sMCI
pMCI
21 (72.7)
12 (75.7)
Sample collected for the study MRI and MEG: MEG power data on each ROI and hippocampal volumes, age, gender, cognitive reserve, neuropsychological testing score, and APOE genotype HLR Train/test method: 75/25 partition pMCI vs sMCI: 100 0.97
Suk et al. (2016) [70] HS
AD
sMCI
pMCI
52 (75.3) 51 (75.2)
56 (75)
43 (75.7)
ADNI-1 MRI and PET: GM, CSF, and intensity measures DW-S2MTL 10-fold cross validation pMCI vs sMCI: 73 NA
Thung et al. (2016) [71] sMCI pMCI 53 (75.7)
60 (75.2)
ADNI-1 MRI: whole-brain GM volume and changes in 4 years of follow-up Mk-SVM 10-fold cross validation pMCI vs sMCI: 78.2 0.84
Vasta et al. (2016) [72] HS
AD
sMCI
pMCI
47 (78.2)
55 (75.9)
89 (75.4)
32 (75.5)
ADNI-1 MRI: hippocampal volumes SVM
Naive Bayes
Neural Network
Train/test method: AD + HS as train set and MCI as test set pMCI vs sMCI:
SVM: 66.1
Naïve Bayes: 65.3
Neural Network: 66.9
NA
NA
NA
Zhang et al. (2016) [73] sMCI
pMCI
62 (75.4)
71 (74.8)
ADNI-1 MRI: whole ventricular tensor-based morphometry SVM 5-fold cross-validation pMCI vs sMCI: 96.7 0.97
Zhang et al. (2016) [74] AD
HS
MCI
194 (NA)
228 (NA)
388 (NA)
ADNI-1 MRI: multivariate hippocampal surface TBM and radial distance AdaBoost Leave-one-out cross-validation pMCI vs sMCI: 77 0.75
Ҫitak-Er et al. (2017) [75] sMCI
pMCI
165 (70.9)
140 (73.1)
ADNI-1 MRI: GM volumes Linear SVM
Polynomial-SVM
LR
10-fold cross-validation pMCI vs sMCI:
Linear-SVM: 73.1
Poly-SVM: 78.7
LR: 76.1
NA
Hojjati et al. (2017) [76] sMCI
pMCI
62 (73)
18 (73.6)
ADNI-1 MRI: connectivity matrix SVM 9-fold cross-validation pMCI vs sMCI: 91.4 0.95
Long et al. (2017) [77] HS
AD
sMCI
pMCI
135 (76.2)
65 (75.6)
132 (75.2)
95 (75.1)
ADNI-1 MRI: whole-brain GM and WM SVM 10-fold cross-validation pMCI vs sMCI:
with GM: 85.9
with WM: 68.7
GM: 0.89
WM: 0.68
Mathotaarachchi et al. (2017) [78] sMCI
pMCI
230 (71.4)
43 (73.2)
ADNI-1 PET: intensity, demographic, and AOPE4 genotype measures RUSRF 10-fold cross-validation pMCI vs sMCI: 84 0.91
Suk et al. (2017) [79] HS
AD
sMCI
pMCI
226 (NA) 186 (NA)
226 (NA)
167 (NA)
ADNI-1 MRI: GM volume CNN 10-fold cross-validation pMCI vs sMCI:
74.8
0.75
Tong et al. (2017) [80] HS
AD
sMCI
pMCI
229 (75.9)
191 (75.3)
129 (74.6)
171 (74.5)
ADNI-1 MRI: global grading biomarker, age, and cognitive measures SVM 10-fold cross-validation pMCI vs sMCI: 81 0.87
Choi et al. (2018) [81] HS
AD
sMCI
pMCI
182 (73.4)
139 (74.3)
92 (70.3)
79 (72.3)
ADNI-2 PET: voxel volumes of FDG and florbetapir (AV-45) images CNN 10-fold cross-validation pMCI vs sMCI: 84.2 0.89
Donnelly-Kehoe et al. (2018) [82] HS
AD
sMCI
pMCI
100 (NA)
100 (NA)
100 (NA)
100 (NA)
ADNI-1 MRI: brain morphometry, demographic data, and MMSE RF
SVM
AB
Train/test method: 75/25 partition NA 0.75
0.76
0.63
Gao et al. (2018) [83] HS
AD
MCI
94 (76.3)
58 (74.2)
147 (74.8)
ADNI-1 MRI and PET: hippocampal textrure features, medical history, and neuropsychological tests GPR
PLS
Train/test method: AD + HS as train set and MCI as test set + follow-up pMCI vs sMCI:
GPR:82.2
PLS:85.5
NA
Gómez-Sancho et al. (2018) [84] HS
AD
sMCI
pMCI
413 (NA)
326 (NA)
173 (NA)
274 (NA)
ADNI-1 MRI: regional volumetry, surface area, and cortical thickness SVM
RLR
10-fold cross-validation pMCI vs sMCI:
SVM: 61-62.5
RLR: 61.1-65
0.64-0.68
0-65-0.70
Hojjati et al. (2018) [85] sMCI
pMCI
62 (73)
18 (73.6)
ADNI-1 MRI: whole-brain cortical thickness, volumes, and connectivity matrix SVM 9-fold cross-validation pMCI vs sMCI: 97 0.98
Khanna et al. (2018) [86] HS
MCI
315 (NA)
609 (NA)
ADNI-1 MRI and PET: volume, clinical, and SNP measures GBM 10 iterations of a 10-fold cross-validation C-index (it is a generalization of the AUC ROC calculation for binary classification): 0.86
Lin et al. (2018) [87] HS
AD
sMCI
pMCI
229 (NA)
188 (NA)
139 (NA)
169 (NA)
ADNI-1 MRI: hippocampal measures and whole-brain cortical volume, surface area, and cortical thickness LASSO + ELM Leave-one-out cross-validation pMCI vs sMCI: 79.9 0.86
Liu et al. (2018) [88] HS
AD
sMCI
pMCI
230 (77.1)
200 (76.6)
160 (76.2)
120 (75.9)
ADNI-1 MRI: whole-brain hierarchical structural network Multiple Kernel Boost 10-fold cross-validation pMCI vs sMCI: 72.8 0.72
Liu et al. (2018) [89] HS
AD
MCI
126 (76)
186 (75.4)
395 (74.9)
ADNI-1 MRI and PET: volume, intensity, and CSF (Aβ42, t-tau y p-tau) measures Multi-hyper graph Learning 10-fold cross-validation pMCI vs sMCI: 74.7 0.72
Lu et al. (2018) [90] HS
AD
sMCI
pMCI
360 (73.4)
238 (75)
409 (74)
217 (74)
ADNI-1 MRI and PET: volume, intensity, and CSF (Aβ42, t-tau y p-tau) measures Deep NN 10-fold cross-validation pMCI vs sMCI: 82.9 NA
Minhas et al. (2018) [91] sMCI
pMCI
54 (74.1)
65 (74.7)
ADNI-1 MRI: whole-brain volumes, surface area, cortical thickness, and neuropsychological measures SVM 5-fold cross-validation pMCI vs sMCI: 84.3 0.89
Popuri et al. (2018) [92] sHS
uHS
pSH
pMCI
sMCI
eDAT
sDAT
360 (75.4)
52 (78.9)
18 (78.2)
205 (74.8)
431 (75)
133 (76.6)
238 (75.8)
ADNI-1 PET: glucose metabolic signal FPDS Independent test set Classification of DAT+/DAT-:
pMCI = 67.9
sMCI = 70.4
pMCI vs sMCI at 2, 3, and 5 years:
0.81
0.80
0.77
Sorensen et al. (2018) [93] HS
AD
sMCI
pMCI
100 (NA)
100 (NA)
100 (NA)
100 (NA)
ADNI-1 MRI: brain volumetry, cortical thickness, WM hipointensities, MMSE, age, and hippocampal sub-regional volumetry Linear SVM
RBF SVM
Train/test method: 60/40 partition pMCI vs sMCI:
Linear SVM: 55.6
RBF SVM: 55
NA
NA
Sun et al. (2018) [94] HS
AD
sMCI
pMCI
162 (76.3)
137 (76)
134 (74.5)
76 (74.8)
ADNI-1 MRI: GM densities LASSO + SVM Train/test method: 50/50 partition pMCI vs sMCI: 65.4 0.68
Wu et al. (2018) [95] HS
sMCI
pMCI
150 (75.6)
150 (75.3)
157 (75.3)
ADNI-1 MRI: 3D brain volumes GoogleNet
CaffeNet
5-fold cross-validation GoogleNet/CaffeNet in: pMCI: 84.7/92.3
sMCI: 67.3/72
NA
Yan et al. (2018) [96] sMCI
pMCI
50 (NA)
29 (NA)
ADNI-1 PET: 3D images ResNet 10-fold cross-validation pMCI vs sMCI: 82 0.81
Basaia et al. (2019) [97] HS
AD
sMCI
pMCI
352 (74.5)
294 (75.1)
510 (72.3)
253 (73.8)
ADNI-1 MRI: WM, GM, and CSF measures CNN Independent test set pMCI vs sMCI: 74.9 NA
Cheng et al. (2019) [98] HS
AD
MCI
112 (NA)
102 (NA)
192 (NA)
ADNI-1 MRI: GM volumes and CSF measures SVM 10-fold cross-validation pMCI vs sMCI: 76.3 0.81
Collazos-Huertas et al. (2019) [99] sMCI
pMCI
325 (75)
245 (75)
ADNI-1 MRI: volumes and cortical thickness SVM
KNN
5-fold cross-validation pMCI vs sMCI:
SVM: 76.1
KNN: 77.8
NA
NA
Elahifasaee et al. (2019) [100] HS
AD
sMCI
pMCI
229 (76)
198 (57.5)
236 (74.9)
167 (74.9)
ADNI-1 MRI: GM density KDA 10-fold cross-validation pMCI vs sMCI: 65.9 0.71
Ezzati et al. (2019) [101] HS
AD
sMCI
pMCI
424 (74.3)
249 (74.7)
372 (72.8)
235 (73)
ADNI-1 MRI: whole-brain volumes, APOE4 genotype, and demographic measures Ensemble Learning 10-fold cross-validation MCI to AD at 6, 12,
24, 36, and 48 months: 63.8%, 68.9%, 74.9%, 75.3%, and 77%, respectively
NA
Gupta et al. (2019) [102] HS
AD
sMCI
pMCI
38 (76.7)
38 (77.1)
36 (74.2)
46 (76.7)
ADNI-1 MRI and PET: whole-brain volume, intensity and CSF (Aβ42, t-tau y p-tau) measures Mk-SVM 10-fold cross-validation pMCI vs sMCI: 94.9 0.94
Lee et al. (2019) [103] HS
AD
sMCI
pMCI
415 (NA)
338 (NA)
558 (NA)
307 (NA)
ADNI-1 MRI: brain phenotypes, demographic and neuropsychological data, APOE4 genotype, and CSF measures rDNN 5-fold cross-validation pMCI vs sMCI at 6, 12, 18 and 24 months: 81, 81, 79 and 80 respectively NA
Lee et al. (2019) [104] HS
AD
sMCI
pMCI
229 (76)
198 (75.4)
214 (75)
160 (74.9)
ADNI-1 MRI: GM volumes rDNN + SVM 10-fold cross-validation pMCI vs sMCI: 88.5 0.96
Lei et al. (2019) [105] HS
AD
sMCI
pMCI
152 (NA)
91 (NA)
98 (NA)
104 (NA)
ADNI-1 MRI: GM volumes and neuropsychological measures SVR 10-fold cross-validation pMCI vs sMCI: 78 NA
Li et al. (2019) [106] NA 803 (NA) ADNI-1/2/GO and AIBL MRI: hippocampal volumes, cognitive, demographic and neuropsychological measures NN Train in ADNI-1 and test in ADNI-2&GO MCI to AD prediction (C-index): 0.86
Li et al. (2019) [107] HS
AD
sMCI
pMCI
165 (76.4)
142 (76.1)
95 (74.9)
126 (73.4)
ADNI-1 MRI: cortical thickness and volumes SVM with RBF kernel 10-fold cross-validation pMCI vs sMCI: 69.8 0.70
Oh et al. (2019) [108] HS
AD
sMCI
pMCI
230 (76)
198 (75.6)
101 (74.1)
166 (74.8)
ADNI-1 MRI data CNN 5-fold cross-validation pMCI vs sMCI: 73.9 NA
Pan et al. (2019) [109] HS
AD
sMCI
pMCI
90 (76.1)
94 (75.8)
44 (77.6)
44 (75.9)
ADNI-1 PET: intensities and connectivity measures SVM 10-fold cross-validation pMCI vs sMCI: 72.3 0.72
Pusil et al. (2019) [110] sMCI
pMCI
27 (71.2)
27 (74.8)
Sample collected for the study MEG: brain connectivity matrix MCFS + SVM with RBF kernel Train/test method: 80/20 partition pMCI vs sMCI: 100 NA
Spasov et al. (2019) [111] HS
AD
sMCI
pMCI
184 (74.6)
192 (75.6) 228 (72.2)
181 (73.7)
ADNI-1 MRI: brain volumes, demographic, neuropsychological, and genetic (APOE4) measures CNN Train/test method: 90/10 partition pMCI vs sMCI: 86 0.93
Wang et al. (2019) [112] HS
AD
MCI
71 (72.5)
48 (75)
60 (72.6)
ADNI-2 MRI: morphometry and WM structural connectivity LR 10-fold cross-validation pMCI vs sMCI: 59 0.65
Wee et al. (2019) [113] HS
AD
MCI
eMCI
lMCI
ADNI-1/ADNI-2:
242/300 (76.9/75.6)
355/261 (76.3/75.3)
415/NA (75.9)
NA/314 (72.9)
NA/208 (73.7)
MRI: cortical thickness Graph NN 10-fold cross-validation Conversion from:
eMCI to AD: 79.2
lMCI to AD: 65.2
NA
Xu et al. (2019) [114] HS
AD
sMCI
pMCI
165 (76.4)
142 (76.1)
95 (74.9)
126 (73.4)
ADNI-1 MRI: cortical thickness SVM with RBF kernel 10-fold cross-validation pMCI vs sMCI: 63.7 0.67
Zhou et al. (2019) [115] HS
AD
sMCI
pMCI
204 (76.1)
171 (75.5)
205 (75.1)
157 (74.8)
ADNI-1 MRI and PET: GM volumes, average intensities and SNP measures Multi-modal Classifier 10-fold cross-validation pMCI vs sMCI: 74..3 0.75
Zhu et al. (2019) [116] HS
AD
MCI
101 (75.8)
93 (75.4)
202 (75.1)
ADNI-1 MRI and PET: GM volumes and average intensities SVM 10-fold cross-validation pMCI vs sMCI: 72.6 0.73
Abrol et al. (2020) [117] HS
AD
sMCI
pMCI
237 (74.3)
157 (75.1)
245 (72.1)
189 (74.2)
ADNI-1
ADNI-2
ADNI-3
ADNI-GO
MRI: 3D brain volumes ResNet Train/test method: 80/20 partition pMCI vs sMCI: 75.1 0.78
Gao et al. (2020) [118] HS
sMCI
pMCI
847 (56.9)
129 (74.8)
168 (74.8)
ADNI-1 MRI: 3D brain volumes Age prediction + AD-NET 5-fold cross-validation pMCI vs sMCI; 76 0.81
Giorgio et al. (2020) [119] HS
MCI
317 (NA)
272 (NA)
ADNI-1 MRI and PET: GM density, genetic, and cognitive measures GMLVQ 10-fold cross-validation pMCI vs sMCI: 81.4 NA
Khatri et al. (2020) [120] HS
AD
MCI
57 (75.6)
53 (74.4)
77 (74.1)
ADNI-1 MRI: cortical thickness,
surface area, GM volumes, MMSE, APOE4 data, and levels of Aβ42, T-tau and P-tau in CSF
SVM with RBFk
Linear SVM
ELM
10-fold cross-validation pMCI vs sMCI:
SVM-RBFk: 71.3
Linear SVM: 75.7
ELM: 83.4
NA
NA
0.85
Lin et al. (2020) [121] HS
AD
sMCI
pMCI
200 (73.9)
102 (75.7)
205 (71.8)
110 (73.9)
ADNI-1 MRI and PET: volume, cortical thickness, intensity measures, APOE4 presence, and levels of Aβ42, T-tau, and P-tau in CSF LASSO + ELM with Gaussian kernel 10-fold cross-validation pMCI vs sMCI: 84.7 0.89
Lin et al. (2020) [122] sMCI
pMCI
124 (70.8)
40 (71.6)
ADNI-1 MRI: GM densities SVM 4-fold cross-validation pMCI vs sMCI: 97.3 0.98
Pan et al. (2020) [123] HS
AD
sMCI
pMCI
262 (NA)
237 (NA)
173 (NA)
115 (NA)
ADNI-1 MRI: 3D brain volumes CNN + EL 5-fold cross-validation on independent test set pMCI vs sMCI: 62 0.59
Ramon-Julvez et al. (2020) [124] HS
AD
sMCI
pMCI
181 (NA)
191 (NA)
227 (NA)
179 (NA)
ADNI-1 MRI data and Jacobian determinant of diffeomorphic transformations CNN 10-fold cross-validation pMCI vs sMCI: 89 0.94
Xiao et al. (2020) [125] HS
AD
sMCI
pMCI
50 (77.8)
51 (75.8)
45 (71.9)
51 (72.5)
ADNI-1 MRI: GM volumes LR 10-fold cross-validation pMCI vs sMCI: 72.9 NA
Xu et al. (2020) [126] HS
MCI
53 (69.6)
76 (73.7)
Sample collected for the study MEG: brain connectivity matrix MG2G Embedding model Train/validation/test method: 85/10/5 partition HS vs pMCI vs sMCI: 82
pMCI vs sMCI: 87
0.75-0.96
Yang et al. (2020) [127] sMCI
pMCI
280 (72)
70 (71.7)
ADNI-1 PET: GM densities CNN + SVM Train/test method: unknown partition pMCI vs sMCI: 78.6 NA
Yee et al. (2020) [128] sHS
uHS
pSH
pMCI
sMCI
eDAT
sDAT
359 (75.4)
51 (79)
19 (78.1)
210 (75)
427 (75)
135 (76.6)
237 (75.7)
ADNI-1 PET: intensity measures CNN 5-fold cross-validation pMCI vs sMCI: 74.7 0.81
Zhou et al. (2020) [129] HS
AD
sMCI
pMCI
226 (75.8)
186 (75.3)
205 (75.1)
157 (74.7)
ADNI-1 MRI and PET: GM volumes and intensity measures SVM 10-fold cross-validation pMCI vs sMCI: 74-76 0.74-0.76
Bae et al. (2021) [130] HS
AD
sMCI
pMCI
2084 (76.5)
1406 (76.2)
222 (72.2)
228 (74.2)
ADNI-1 MRI: 3D brain volume and neuropsychological measures CNN Train/validation/test method: 70/15/15 partition pMCI vs sMCI: 82.4 NA
Mofrad et al. (2021) [131] sMCI
pMCI
333 (NA)
134 (NA)
ADNI-1 MRI: hippocampal entorhinal cortex, ventricles, and neuropsychological measures Ensemble Learning 15-fold cross-validation pMCI vs sMCI: 77 NA
Mofrad et al. (2021) [132] sMCI
pMCI
279 (NA)
279 (NA)
ADNI-1 and AIBL MRI: hippocampal and ventricle measures Ensemble Learning 15-fold cross-validation pMCI vs sMCI: 78 NA
Pan et al. (2021) [133] HS
AD
sMCI
pMCI
242 (73.6)
237 (75)
360 (71.7)
166 (73.9)
ADNI-1 PET: 3D images CNN 5-fold cross-validation repeated 2 times pMCI vs sMCI: 83 0.87
Shen et al. (2021) [134] HS
AD
sMCI
pMCI
150 (NA)
143 (NA)
89 (NA)
86 (NA)
ADNI-1 MRI and PET: Volume, cortical thickness, intensity measures, APOE4 presence and levels of Aβ42, T-tau, and P-tau in CSF SVM 10-fold cross-validation pMCI vs sMCI: 75-78 0.76-0.80
Syaifullah et al. (2021) [135] HS
AD
MCI
543 (NA)
359 (NA)
544 (NA)
NA-ADNI MRI and PET data and MMSE SVM Train/test method: 50/50 partition pMCI vs sMCI: 87.9 NA
Wen et al. (2021) [136] HS
AD
MCI
sMCI
pMCI
46 (72.7)
46 (74.4)
97 (72.9)
54 (72.6)
24 (74.2)
ADNI-1 MRI: GM density SVM 10-fold cross-validation pMCI vs sMCI: 80 NA
Zhang et al. (2021) [137] HS
AD
sMCI
pMCI
275 (76.2)
280 (76.1)
251 (77.6)
162 (75.1)
ADNI-1 MRI: 3D brain volumes CNN Train/validation/test method: 70/15/15 partition pMCI vs sMCI: 78.8 0.87
Zhu et al. (2021) [138] HS
AD
sMCI
pMCI
275 (76.2)
280 (76.1)
251 (77.6)
162 (75.1)
ADNI-1 MRI: GM, WM, and CSF measures, demographic data and APOE genotype Temporally structured-SVM 10-fold cross-validation pMCI vs sMCI: 85.4 0.86
  1. Note. AB Ada-Boost, AD Alzheimer’s disease, AD-NET Age-adjust neural network, AIBL Australian Imaging, Biomarkers and Lifestyle Flagship Study of Aging, ANN Artificial neural network, AUC Area under the curve, CNN Convolutional neural network, DAT Dementia Alzheimer type, DBM Deep Boltzmann Machine, DW-S2MTL Deep-weighted subclass-based sparse multi-task learning, EL Ensemble learning, eDAT Early DAT, ELM Extreme learning machine, eMCI Early MCI, EN Elastic nets, F-FDG Fluorine 18 fluorodesoxyglucose, FPDS FDG-PET, GM Gray matter, GMB Gradient boosting model, GPR Gaussian process regression, HS Healthy subjects, HLR Hierarchical logistic regression, ICA Independent component analysis, KDA Kernel discriminant analysis, lMCI Late MCI, LR Logistic regression, M2TL Multimodal manifold-regularized transfer learning, M3TL Multi-modal multi-task learning, MCI Mild cognitive impairment, MCFS Multi-cluster feature selection, MG2G Multiple Graph2Gauss, MKL Multiple kernel learning, MMSE Mini Mental State Examination, MTSRC Multi-task sparse representation classifier, NA Not applicable, nl-SVM-RBFk Non-linear SVM with radial basis function kernel, NN Neural network, OPLS Orthogonal partial least squares, PBL-McqRBFN Projection-based learning for meta-cognitive q-Gaussian radial basis function network, PLS Partial least squares, pMCI Progressive MCI, rDNN Randomized deep neural network, Res-Net deep residual neural network, RF Random forest, RLR Regularized logistic regression, RUSRF Random under sampled random forest, sDAT Stable DAT, SM2TLC Sparse multimodal manifold-regularized transfer learning classification, sMCI Stable MCI, SNN Spiking neural network, SNP Single-nucleotide polymorphisms, ss Sample selection, SVM Support vector machine, VFI Voting feature intervals, WM White matter, wmSRC Weighted multi-modality sparse representation-based classification