Table 3.
Datasets, data analysis methods, and key findings for selected studies in research foci 1 - diagnosing AD or MCI.
| Study | Datasets | Primary data analysis methods | Key findings |
|---|---|---|---|
| Van Gils et al. (23) | ADNI data (229 HC, 402 MCI, 190 AD) and data from Kuopio L-MCI study (687 HC, 249 MCI, 77 AD) | Support vector machine (SVM) and linear model. Performance of separating persons with AD from those with MCI or HC using combined features is 94%–100%. | Identification of efficient biomarker sets (including Apolipoprotein E (ApoE) alleles, CSF, estrogen usage duration, cognitive and memory tests and MRI features) related to AD diagnosis |
| Li et al. (24) | MRIs from ADNI1 (80 AD, 141 MCI, 142 HC) | Support vector machine with selected imaging features from structured MRI. | Selected AD-specific anatomical features from structured MRI have discriminative capability in differentiating AD or MCI from healthy controls. |
| Yang et al. (25) | Clinical dementia rating, MMSE and MRI scans from 17 AD, 18 MCI, 17 HC | Support vector machine with particle swarm optimization (PSO) and principle component analysis (PCA). Diagnosis accuracy: 94% for AD and 88.9% for MCI | SVM-PSO with PCA can classify AD and MCI versus HC |
| Mangialasche et al. (26) | Structural MRI measures, plasma levels of vitamin E and makers of vitamin E oxidative/nitrosative damage (81 AD, 86 MCI and 86 HC from AddNeuroMed study) | Multivariate data analysis (Orthogonal partial least squares to latent structures (OPLS)), with 67 variables from structural MRI measures and plasma levels of vitamin E forms. | Plasma levels of tocopherols and tocotrienols with MRT can differentiate AD and MCI from HC subjects and predict MCI to AD conversion |
| Kohannim et al. (27) | MRI and biomarkers from ADNI (158 AD, 366 MCI, 213 HC) | Support vector machine with brain imaging and other biomarkers features | SVM with brain imaging and biomarkers can classify AD, MCI and HC. |
| Clark et al. (28) | Semantic fluency word lists, dementia rating, and neuropsychological assessment (training set: 41 AD, 80 MCI, 44 HC; testing set: 9 AD, 21 MCI, 35 HC) | Random forest classifier | Semantic fluency lists can potentially predict functional declines |
| Lagun et al. (29) | Eye movement data and neuropsychological assessment (20 AD, 10 MCI, 30 HC) | Naïve Bayes, Logistic regression, Support vector machine | Eye movement measures with SVM classification techniques can detect MCI |
| Casanova et al. (30) | Structural MRI, DNA, and cognitive data of Caucasians in ADNI (171 AD, 153 PMCI, 182 SMCI, 188 CN) | Regularized logistic regression (RLR) | A new metrics, AD pattern similarity (AD-PS) scores, was designed to assess risk of AD. |
| Zhang et al. (32) | MRI, FDG-PET, CSF data from ADNI (45 AD, 91 MCI, 50 HC) | Multi-modal multi-task (M3T) learning | M3T learning performed well on both AD detection and clinical score prediction |
| Mattila et al. (33) | ADNI (163 AD, 190 SMCI, 154 PMCI, 199 HC) | Disease state fingerprint visualization, statistical disease state index (DSI) method | DSI can estimate AD state |
Note: ADNI, Alzheimer’s Disease Neuroimaging Initiative database; MCI, mild-cognitive impairment; SMCI, stable mild cognitive impairment; PMCI, progressive mild cognitive impairment; HC, healthy control.