The inclusion criteria were as follows: 1) diagnosis of probable AD; 2) initial Montreal Cognitive Assessment (MOCA) score of 5–26. All patients were required to meet the diagnostic criteria for the diagnosis of mild cognitive impairment due to Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease.

    The inclusion criteria were as follows: (1) Concern regarding a change in cognition; (2) Impairment in one or more cognitive domains; (3) Preservation of independence in functional abilities; (4) Not demented.The exclusion criteria were: (1) Parkinsonism, including prominent visual hallucinations, and rapid eye movement sleep abnormalities, often seen in dementia with Lewy bodies; (2) multiple vascular risk factors and/or the presence of extensive cerebrovascular disease on structural brain images, which is suggestive of vascular cognitive impairment;(3) prominent behavioral or language disorders early in the course of disease that may reflect frontotemporal lobar degeneration, or (4) very rapid cognitive decline that occurs over weeks or months, typically indicative of prion disease, neoplasm, or metabolic disorders.

    EEG recording: EEG data were recorded from 64 channels with a BrainAmp DC amplifier (Brain Products GmbH, Germany) under eyes-closed conditions in a metallic-shielded room. FCz was used as the online reference and AFz as the ground. Data were initially sampled at 5000 Hz with impedances kept below 5KΩ for all channels. Prior to EEG collection, all participants were asked not to consume any caffeine or energy-related drinks for up to 24 hours.

    EEG preprocessing: Semi-automated EEG pre-processing for artifact rejection was performed using the EEGLAB toolbox and ICLabel toolbox. The preprocessing steps are briefly described as follows. (1) Resampled the raw data to 250 Hz; (2) Visually inspected the EEG signals and manually removed bad segments contaminated by movement artifacts; (3) Applied bandpass filter of 0.5-100 Hz and notch filter of 50 Hz to the down-sampled signals; (4) Discarded bad channels by checking the spectra of all channels. The rejected bad channels were then interpolated using neighboring channels via spherical spline interpolation; (5) Removed the remaining artifacts using independent component analysis (ICA). In this step, ICs marked by ICLables as eye movement, muscle, heart, channel noise and line noise were rejected, and ICs labelled as other would be checked by the operators manually to decide to keep or reject; (6) EEG data were re-referenced to the common average. (7) The resulting EEG data were finally visually inspected again for any possible remaining noise.