Characteristic indicators for mild-cognitive-impairment obtained from dimension reduction of brain networks

Alzheimer's Disease (AD) is irreversible. Mild cognitive impairment (MCI) is the first symptomatic stage of AD. Distinguishing MCI patients from healthy controls (HCs) through appropriate techniques is critical for early therapeutic interventions and prolonging patients' health. In this study, we explored characteristic indicators for MCI through dimension reduction of brain networks. After the brains (100 HCs and 100 MCIs from ADNI dataset) were partitioned into 360 parcels, one-dimension time series was extracted from diffusion and resting-state functional magnetic resonance imaging (MRI) data using network dimension-reduction techniques. Power-spectrum was then employed to transform the time series into frequency domain, to find characteristic indicators for MCI. Statistical tests indicate that the indicators (mean square frequency, and center frequency) estimated with brain network reductions could differentiate MCIs from HCs more significantly, compared with those of BOLD time series of specific AD-related subregions (hippocampus, and parieto-temporal subregions). Power-spectrum of one-dimension time series extracted with network reductions might be a viable method for distinguishing MCI progression stages. This approach could potentially facilitate earlier and more precise differentiation between MCIs and HCs, showing future clinical applicability.