We argue that whole-brain computational models are well-placed to achieve this objective. In line with our dynamic hypothesis, we suggest that aging needs to be studied on a continuum rather than at discrete phases such as ‘adolescence’, ‘youth’, ‘middle-age’, and ‘old age’. We propose that these significant epochs of the lifespan can be related to shifts in the dynamical working point of the system. We recommend that characterization of metastability (wherein the state changes in the dynamical system occur constantly with time without a seeming preference for a particular state) would enable us to track the shift in the dynamical working point across the lifespan. This may also elucidate age-related changes in cognitive performance. Thus, the changing structure–function–cognition relationships with age can be conceptualized as a (new) normal response of the healthy brain in an attempt to successfully cope with the molecular, genetic, or neural changes in the physiological substrate that take place with aging, and this can be achieved by the exploration of the metastable behavior of the aging brain.The authors proceed to illustrate structural and functional connectivity changes during aging, as white-matter fiber counts reduce, roles of hub, feeder and local connections change, and brain function becomes less modular. I want to pass on their nice description of the healthy aging brain:
Age differences in cognitive function have been studied to a great extent by both longitudinal and cross-sectional studies. While some cognitive functions − such as numerical and verbal skills, vocabulary, emotion processing, and general knowledge about the world − remain intact with age, other mental capabilities decline from middle age onwards: these mainly include episodic memory (ability to recall a sequence of events as they occurred), processing speed, working memory, and executive control. Age-related structural changes measured by voxel-based morphometry (VBM) studies have reported expansion of ventricles, global cortical thinning, and non-uniform trajectories of volumetric reduction of regional grey matter, mostly in the prefrontal and the medial temporal regions. While the degeneration of temporal–parietal circuits is often associated with pathological aging, healthy aging is often associated with atrophy of frontostriatal circuits. Network-level changes are measured indirectly by deriving covariance network of regional grey matter thickness or directly by diffusion weighted imaging methods which can reconstruct the white matter fiber tracts by tracking the diffusion of water molecules. These studies have revealed a linear reduction of white matter fiber counts across the lifespan. The hub architecture that helps in information integration remains consistent between young adults and elderly adults, but exhibits a subtle decrease in fiber lengths of connections between hub-to-hub and non-hub regions. The role of the frontal hub regions deteriorates more than that of other regions. The global and local measures of efficiency show a characteristic inverted U-shaped curve, with peak age in the third decade of life. While tractography-based studies report no significant trends in modularity across the lifespan, cortical network-based studies report decreased modularity in the elderly population. Functional changes derived from the level of BOLD signal of the fMRI during task and rest (i.e., in the absence of a task) demonstrate more-complex patterns such as task-dependent regional over-recruitment or reduced specificity. More interesting changes take place in functional networks determined by second-order linear correlations between regional BOLD time-series in the task-free condition. Modules in the functional brain networks represent groups of brain regions that are collectively involved in one or more cognitive domains. An age-related decrease in modularity, with increased inter-module connectivity and decreased intra-module connectivity, is commonly reported. Distinct modules that are functionally independent in young adults tend to merge into a single module in the elderly adults. Global efficiency is preserved with age, while local efficiency and rich-club index show inverted U-shaped curves with peak ages at around 30 years and 40 years, respectively. Patterns of functional efficiency across the cortex are not the same. Networks associated with primary functions such as the somatosensory and the motor networks maintain efficiency in the elderly, while higher-level processing networks such as the default mode network (DMN), frontoparietal control network (FPCN), and the cingulo-opercular network often show decline in efficiency. Any comprehensive aging theory requires an account of all these changes in a single framework.