Phase aligned EEG source space features for distinguishing Alzheimer’s disease from frontotemporal dementia

Alzheimer’s disease (AD) and frontotemporal dementia (FTD) share similarities in cognitive deficits and behavioural changes, increasing the difficulty of their differential diagnosis. Electroencephalography (EEG), as a non-invasive, cost-effective and easy-to-use tool that reflects brain activity in real time, provides potential support for differentiating AD and FTD. We propose a source-space feature method based on EEG phase alignment, where dynamic time warping (DTW) enhances inter-trial phase synchrony in the θ, α, β, and γ bands, sLORETA extracts brain region energy features, and a random forest classifier performs classification. In a dataset comprising 36 AD cases, 23 FTD cases, and 29 CN, the classification accuracies for AD vs. FTD and FTD vs. CN reached 94.21% and 90.63%, respectively. Abnormalities in the AD group were primarily observed in the θ and γ bands, involving multiple cognition-related regions, whereas the FTD group exhibited β-band abnormalities predominantly affecting the parietal lobe and default mode network. These findings suggest that source-space features based on phase alignment can effectively distinguish the neural signatures of AD and FTD.