Eric Kwun Kei Ng1, Siwei Liu1, Yijun Wang1, Jia-Hou Poh1, Yng Miin Loke1, Boon Linn Choo1, Beatrice Rui-Yi Loo1, Joseph Lim1, Hee Youn Shim1, Lingjie Zhu1, Joanna Su Xian Chong1, Chenhao Wang1, Juan Zhou1,2
1Centre for Cognitive Neuroscience, Duke-NUS Medical School, Singapore, 2A*STAR-NUS Clinical Imaging Research Centre, Singapore
Brain functional networks show high similarity and subtle differences at rest and task . Accumulating evidence suggests that intrinsic connectivity at rest provides the communication channels of task information ; high similarity between the rest and task architectures may indicate less network reconfigurations and better cognitive capacity . Based on this framework, networks during task may become more integrated to handle higher cognitive demands . Previous studies suggested aberrant network organization such as less segregated associative networks at rest in elderly . Whether the ageing brain  requires more task-related alternations in connectivity for competent functioning remains to be fully elucidated. Here, we examined such connectivity changes in three age groups during rest and a spatial working memory task.
20 adolescents (13-18 years), 20 adults (19-26), and 16 healthy elderly (55-70) performed an 8-10-minute eyes-open task-free fMRI followed by 4 runs of spatial working memory task in a 3T scanner (TR = 2s) . They had to remember the spatial locations of 6 dots (simultaneously presented for 800ms) for 14 seconds. A face (800ms) appeared halfway of the retention period, creating 2 levels of cognitive demands: distraction (D), when they viewed the face passively; and the more demanding interference (I), when they judged and responded if it was a target. After retention, they had to decide if the location of a probe (800ms) matched any of the 6 dots within 4.8 seconds. Task order was counterbalanced (IIDD or DDII; 18 trials per run).
Task accuracy and response time were analyzed with generalized linear mixed models. fMRI data went through standard preprocessing without global signal regression; task activation was removed for task data . Static functional connectivity matrices (FC; Fisher z-transformed Pearson correlation) were obtained for 430 ROIs  with additional subcortical ROIs. FC matrices were analyzed with partial least square , which identified latent variables (LVs) best capturing the covariance between FC and age group. 1000 permutation and 500 bootstrapping were computed to assess statistical significance of an LV (p < .05) and its most salient FC edges, respectively.
PLS was first conducted on mean task FC to identify general age differences in task architecture. Its task-rest similarity (spatial Pearson correlation) was then compared across groups and correlated with performance. Another PLS was done on task difference FC to probe age-related FC changes due to task demand.
Groups did not differ in accuracy (mean 72%, SD 22%), but elderly responded slower than the others (p < .001). PLS on task mean FC yielded 2 LVs, featuring age-related increase in the FC between the default mode (DMN) and task-positive networks (Fig. 1 left), and stronger FC within and between task-positive networks in adults than the others (Fig.1 right). Task-rest similarity was the lowest in elderly (mean r = .59, p < .001) while comparable between adolescents and adults. Across all participants, lower similarity was associated with slower response (r = -.31 p = .018) and interestingly better overall accuracy (r = -.3, p = .03).
PLS on task difference FC yielded 1 LV featuring increase in FC within the DMN and control network, and between DMN/control and task-positive networks in elderly compared to adults during interference (Fig. 2). Stronger LV was correlated with slower response (r = .39, p = .003).
·Figure 1. Older brain involved higher FC between the default mode and task-positive networks and lower FC among task-positive networks during task
·Figure 2. Elderly needed more FC alternations for more difficult condition (interference)
Adult and adolescent task FC was more similar than with the elderly. Compared to adults, elderly FC featured higher coupling between DMN and task-positive networks, more changes to increased task demand, and more task-rest reconfigurations. Ageing brain may be less efficient such that it requires extra reorganization according to various cognitive demands, i.e., more coupling across networks . These extra efforts might have allowed elderly to perform adequately , but at the expense of speed.
BOLD fMRI 2
Other - functional connectivity, task vs. rest
1|2Indicates the priority used for review