Associations between a biological marker of resilience, brain pathology, and cognitive functioning

Abstract

AbstractBackgroundDementia is the end‐product of decades of accumulating neuropathology. Resilience mechanisms can help sustaining the brain structure (brain maintenance, BM) and/or compensating for neuropathology (cognitive reserve, CR), preserving cognition. Measuring such mechanisms, however, has been challenging with traditional approaches as they fail to capture the complex biological dimension. Using deep learning (DL), we recently developed a biological measure of brain age from raw brain MRI images. Here, we tested whether differences between predicted brain age and the chronological age (PBA‐CA) can be used as marker of BM/CR following the latest framework from the NIH‐supported Collaboratory on Research Definitions for Reserve and Resilience.MethodWithin the Gothenburg H70‐Birth cohort 1944, we included 728 septuagenarians with available brain MRI and without cognitive impairments/neuropsychiatric disorders. We applied the DL algorithm (developed in 15115 healthy individuals from four other cohorts) to the participants’ MRI images to predict brain age and compute PBA‐CA, which was categorized in 4 groups according to distribution: Q1 (younger‐appearing brain) to Q4 (older‐appearing brain). Brain pathology included measures of Alzheimer‐neurodegeneration (average cortical thickness in signature areas), small vessel disease (SVD, i.e. presence of lacunes, microbleeds, white‐matter lesions, perivascular spaces), and microstructural white‐matter changes. A global cognitive score was generated by averaging z‐scores from ten tests. Sex‐adjusted regressions were used in data analysis.ResultParticipants with younger‐appearing brains (reference=older‐appearing) were more likely to have less SVD burden (reference: ≥2 SVD; OR=2.75 [95% CI 1.41–5.37] for no‐SVD; OR=2.55 [95% CI 1.20–5.45] for one‐SVD), white‐matter microstructural alterations, or AD‐related atrophy. Younger‐appearing brains were also associated with better cognitive performance (β‐coefficient=0.19 [95% CI 0.04–0.30]).ConclusionPreliminary findings indicate that negative differences between PBA and CA (PBA<CA) were associated with less AD‐ and cerebrovascular pathology, and better cognitive functioning. This suggests that PBA<CA indicate preserved brain structure and could be used as marker of BM.