Longitudinal brain age prediction and cognitive function after stroke

Abstract Background Advanced age and poor brain health have been associated with higher risk for more severe clinical and cognitive outcomes following stroke, but more accurate models for clinical prediction are needed. Machine learning based on brain scans can be used to estimate brain age of individual patients, and the corresponding difference from chronological age, the brain age gap (BAG) has been investigated in a range of clinical conditions. Yet, the predictive value for post stroke NCD has not been established. To this end, using longitudinal data we aimed to investigate the association between BAG and post-stroke neurocognitive disorder (NCD) over time. Methods 269 stroke survivors (55.4% women, mean (SD) age = 71 (11) years) were included from the Norwegian Cognitive Impairment After Stroke (Nor-COAST) study. MRI and clinical data were collected shortly after the acute stroke (neuropsychological data first after 3 months) and at 18- and 36 months follow-up. The brain age model was trained in an independent sample using established protocols based on machine learning and brain structural features from Freesurfer and applied to the current dataset. We used linear mixed effects (LME) models and survival analysis to assess the associations between cognitive status and longitudinal brain age. Results LMEs revealed a main effect of BAG on post stroke NCD across time, confirming our main hypothesis that a younger appearing brain may protect against post stroke NCD. For patients with no NCD at baseline, survival analysis suggested that higher baseline BAG was associated with higher risk of post stroke NCD at 18 and 36 months. Conclusion A younger appearing brain is associated with lower risk of post stroke NCD up to 36 months after a stroke, even among those showing no evidence of impairments 3 months after hospital admission. Brain age prediction based on brain scans provides a reliable assessment of brain structure and is sensitive to post stroke NCD with predictive value for cognitive decline among patients with no impairment at the initial assessment.