Multimodal neuroimaging discrimination of Alzheimer’s disease, mild cognitive impairment, and late-life depression using electroencephalography and functional near-infrared spectroscopy: integrating electrophysiological and hemodynamic biomarkers

Objective To identify electrophysiological and hemodynamic characteristics of the cerebral cortex during the resting-state that could help differentiate Alzheimer’s disease (AD), mild cognitive impairment (MCI), and late-life depression (LLD) and integrate these characteristics into a diagnostic model. Methods We recorded oxygenated hemoglobin concentration (HbO) signals detected by functional near-infrared spectroscopy (fNIRS) from the prefrontal cortex, partial parietal cortex, and temporal lobe cortex, as well as electrophysiological signals detected by electroencephalography (EEG). The recording time was 30 min. Then, we used machine learning modeling with the support vector machine (SVM) algorithm to evaluate the diagnostic performances of EEG-based, fNIRS-based, and EEG plus fNIRS-based models for distinguishing AD, MCI, and LLD. Results We investigated the differential neural signatures of patients with AD ( n = 61), MCI ( n = 28), and LLD ( n = 27) using an EEG power spectral analysis across six frequency bands: delta (1–4 Hz), theta (4–8 Hz), alpha (8–13 Hz), beta (13–30 Hz), low gamma (30–45 Hz), and high gamma (55–80 Hz). Two key frequency bands significantly differed among the groups. The alpha band power was significantly higher in the LLD group than in the AD and MCI groups ( p < 0.05). The high gamma power was also significantly higher in the LLD group than in the MCI group ( p < 0.05). Regarding fNIRS, HbO power significantly differed in 11 channels (channels 19, 22, 23, 24, 26, 27, 28, 32, 41, 42, and 43); the values were significantly lower in the AD group than in the MCI group and significantly higher in the MCI group than in the LLD group. The accuracies of the EEG, fNIRS, and combined SVM models were 0.5246, 0.5246, and 0.6066, respectively. Conclusion These findings highlight distinct EEG spectral patterns in patients with LLD compared to those with AD or MCI, particularly in alpha and high-gamma oscillations. These differences could be potential biomarkers for differentiating these conditions. Combining EEG and fNIRS analyses may further elucidate the neurophysiological mechanisms underlying these disorders.