Brain age gap and decline in specific cognitive domains after stroke

Abstract

Introduction

Advanced age is associated with poorer prognosis after stroke. Machine learning based on brain imaging can be used to estimate the age of a patient and compute the difference to chronological age; the brain age gap (BAG). Higher BAG (an older appearing brain) has been reported in a range of clinical conditions and is associated with risk of dementia and cognitive decline following stroke, but the associations and predictive value of brain age prediction for post- stroke decline in specific cognitive domains remain unknown. To this end, using longitudinal data after stroke we tested the hypothesis that higher BAG during the acute stroke phase is associated with decline in specific cognitive domains 3 and 18 months after stroke.

Methods

272 stroke survivors (age = 71.1 (11.2), women = 43.4%) were included from the ‘Norwegian Cognitive Impairment After Stroke (Nor-COAST) study. Clinical and MRI data was collected in the acute phase of the stroke and cognitive assessment of attention, executive function, language, and memory was collected at 3 and 18-months follow-up. FreeSurfer anatomical segmentation machine learning was used to predict age of each patient. Z-scores of the cognitive data normalized by mean and standard deviation of published normative data were calculated.

Results

Mean (SD) z-scores at 3 and 18 months respectively were, for attention; -.8 (2.5) and -.4 (2.2), for executive function; -.6 (1.4) and -4. (1.4), for language; -.6 (1.2) and -.3 (1.5), and for memory; -.8 (1.4) and -.8 (1.2). Generalised linear model analyses revealed significant associations between baseline BAG and all of the cognitive domains at both 3 and 18 months, with low/moderate effect sizes (see figure).

Discussion

We found that higher BAG during the acute phase of a stroke is associated with poorer 3- and 18-month cognitive outcome in attention, executive function, language, and memory.

These findings suggest that BAG may be used as a predictive marker for decline in specific cognitive domains after stroke.