Towards Regional Brain Age Models: Region-Sensitive Morphological Associations of Age and Gender

Background: Biological clocks have enormous potential in healthcare, both for detecting pathological ageing and for evaluating the outcomes of therapeutic or pharmaceutical interventions. They also help promote healthier behaviors in at-risk individuals. For similar reasons, neuroimaging-based brain age measures have received growing interest. Previous research has shown that the brain age can be accurately estimated through many structural MRI modalities, including T1 images. However, the expert knowledge on how the morphological brain measures change over the lifespan remains scattered, which presents difficulties for optimizing the brain age models, for example, in constructing regional brain age models. Towards this end, the present study aims to find out how major demographical variables (namely, age and gender) are associated with the morphological measures extracted over the larger brain subdivisions, especially within the context of Hong Kong population.
Methods: 111 participants aged 18-81 underwent MRI scans, during which T1 MPRAGE images were acquired. Linear mixed-effects models were fitted to examine the influence of age, gender, and lobe (frontal, parietal, temporal, occipital, and the limbic lobe—which is further sub-divided into hippocampus and cingulate cortex) on four cortical morphological measures (cortical volume, area, thickness, and mean curvature) extracted using FreeSurfer.
Results: All cortical volume, area, and thickness were negatively correlated with age (p < .001); however, only the cortical volume and thickness showed region-sensitive ageing trajectories, with the hippocampus showing the slowest reduction among all lobes. In contrast, the mean curvature was found to exhibit an age-related increase (p < .05), but the rate was larger for frontal, parietal, and temporal lobes (p < .001) than for occipital lobe, cingulate, and hippocampus (n.s.). Surprisingly, the positive association with age was only found in female (p < .001) but not in male (p = .99).
Significance: The region-sensitive morphological associations with age and gender reported above will find use in optimizing the brain age models, particularly in translating the traditional global brain age estimates to their regional counterparts.
Acknowledgment: This work is supported by Projects of Research Institute of Smart Ageing (P0038995) awarded to WS-Y. Wang (PI) and MC-M. Fong (Co-PI).