Table 1.
Clinical characteristics of the cohort.
AD: Alzheimer’s disease; aMCI: amnestic MCI; oMCI: other MCI; SCI: subjective cognitive impairment; BZD: benzodiazepine.
Fig 1.
Illustration of multi-channel (N = 4, D = 2) EEG signal modeling with HMM.
Table 2.
The computed epoch-based entropy and bump model features for each subject.
Fig 2.
Box plots of the best features discriminating SCI patients from AD patients.
The figures follow the ranking in order of decreasing relevance: (a) EpEn on all electrodes [8–12] Hz; (b) EpEn on temporal region [8–30] Hz; (c) BM on Frontal region [4–8] Hz; (d) EpEn on all electrodes [8–30] Hz; (e) EpEn on frontal + occipital region [8–30] Hz.
Table 3.
Best combination of features for discriminating SCI patients from AD patients (SCI vs. AD), SCI patients from those with MCI or other pathologies (SCI vs. Other), and AD patients from those with MCI or other pathologies (AD vs. Other).
Fig 3.
Box plots of the most relevant features for discriminating possible AD patients from “Other” patients (patients with MCI or other pathologies).
Figures follow the same order given by the OFR algorithm as noted in Table 3: (a) EpEn on Temporal region [8–30] Hz; (b) EpEn on all electrodes [8–30] Hz; (c) EpEn on Frontal + Occipital region [8–30] Hz; (d) EpEn on all electrodes [12–30] Hz; (e) BM on Temporal region [12–30] Hz; (f) EpEn on all electrodes [8–12] Hz; (g) BM on Frontal region [12–30] Hz.
Table 4.
Confusion matrix for differential AD diagnosis with three groups of patients.
Table 5.
Confusion matrix for differential AD diagnosis with four groups of patients.
Table 6.
Distribution of the misclassified patients in the four groups.
Refer to Table 1 that describes the cohort in details.