A model-based approach for determining attention deficits in mild cognitive impairment using scalp EEG.

Mild cognitive impairment (MCI) is a neurocognitive disorder found in

~30 % of the elderly population. Deficits in visuospatial and executive

functions are common in MCI. However, it is still unclear whether these

deficits are to due to dysfunction in selective attention. This is in part due

to a lack of neural indexes that track early selective attention function in

MCI. Here, we adopted a machine learning approach that uses the inverted

encoding model (IEM) to reconstruct spatially selective neural

representations of visuospatial attention based on slow-going EEG

oscillations at ~8-13Hz, known as the alpha band activity. This model-based

approach allowed us to quantify the strength and precision of attentional

fields in individual subjects with millisecond resolution. Specifically, we

measured behavioral and EEG responses from neurologically healthy

adults (20-59 years old) and elderly individuals with and without MCI (60-72

years old, age-matched), while they performed the attention-cueing Eriksen

Flanker task. Using the IEM, we found that alpha-based reconstructions of

visuospatial attention were significantly weaker with significantly slower

onsets in healthy individuals, whose ages were above 30 years old,

compared to those in a younger group (20-30 years old). That said, spatial

tuning profiles of alpha-based spatial reconstructions remained the same 

throughout late adulthood (31-72 years old). Importantly, MCI patients

showed significantly broader alpha-based spatial reconstructions with

significantly slower onsets, compared to those in the healthy aging group.

Together, these results suggest that MCI is in part contributed by the

diminishing fidelity of selective sensory information processing. Moreover,

this model-based approach could be applied further in future studies to

study attention deficits in different MCI subtypes.