Elsevier

NeuroImage

Volume 148, 1 March 2017, Pages 364-372
NeuroImage

Increased sensitivity to age-related differences in brain functional connectivity during continuous multiple object tracking compared to resting-state

https://doi.org/10.1016/j.neuroimage.2017.01.048Get rights and content

Highlights

  • Age-related changes in cognitive abilities vary between cognitive domains.

  • This is mirrored in differential vulnerability of functional network connectivity.

  • fMRI data was obtained during rest and two load levels of a continuous tracking task.

  • We used machine learning to classify task and load.

  • Task data strongly increased sensitivity to age, compared to rest.

Abstract

Age-related differences in cognitive agility vary greatly between individuals and cognitive functions. This heterogeneity is partly mirrored in individual differences in brain network connectivity as revealed using resting-state functional magnetic resonance imaging (fMRI), suggesting potential imaging biomarkers for age-related cognitive decline. However, although convenient in its simplicity, the resting state is essentially an unconstrained paradigm with minimal experimental control. Here, based on the conception that the magnitude and characteristics of age-related differences in brain connectivity is dependent on cognitive context and effort, we tested the hypothesis that experimentally increasing cognitive load boosts the sensitivity to age and changes the discriminative network configurations. To this end, we obtained fMRI data from younger (n=25, mean age 24.16±5.11) and older (n=22, mean age 65.09±7.53) healthy adults during rest and two load levels of continuous multiple object tracking (MOT). Brain network nodes and their time-series were estimated using independent component analysis (ICA) and dual regression, and the edges in the brain networks were defined as the regularized partial temporal correlations between each of the node pairs at the individual level. Using machine learning based on a cross-validated regularized linear discriminant analysis (rLDA) we attempted to classify groups and cognitive load from the full set of edge-wise functional connectivity indices.

While group classification using resting-state data was highly above chance (approx. 70% accuracy), functional connectivity (FC) obtained during MOT strongly increased classification performance, with 82% accuracy for the young and 95% accuracy for the old group at the highest load level. Further, machine learning revealed stronger differentiation between rest and task in young compared to older individuals, supporting the notion of network dedifferentiation in cognitive aging. Task-modulation in edgewise FC was primarily observed between attention- and sensorimotor networks; with decreased negative correlations between attention- and default mode networks in older adults. These results demonstrate that the magnitude and configuration of age-related differences in brain functional connectivity are partly dependent on cognitive context and load, which emphasizes the importance of assessing brain connectivity differences across a range of cognitive contexts beyond the resting-state.

Introduction

Individual life-span trajectories in brain and cognition are shaped by a dynamic interplay between genetic and environmental factors (Lindenberger, 2014). Aging confers a notable increase in inter-individual variability in cognitive functions (Buckner, 2004, Singh-Manoux et al., 2012), and developing sensitive and specific in vivo biomarkers for identification of individuals at risk for cognitive impairment and dementia is a key challenge for clinical neuroscience.

The brain is intrinsically organized into neuronal networks or nodes of which continuous cross-talk enables cognition and complex behaviors (Sporns et al., 2004). Functional connectivity (FC) measured by functional magnetic resonance imaging (fMRI) refers to the synchronization between nodes, reflecting an organizational principle of the brain (Van Den Heuvel and Pol, 2010). In line with this system-level view of brain function, aging studies converge on less specialized and segregated brain networks, supporting the theory of cognitive and neuronal dedifferentiation in aging (Baltes and Lindenberger, 1997; Chan et al., 2014; Daselaar et al., 2005; Lindenberger, 2014; Park et al., 2012).

Age-related FC differences have been extensively studied, implicating a range of networks including the executive control (ECN), dorsal attention (DAN), default mode (DMN), frontoparietal (FPN) and hippocampal networks (Salami et al., 2014, Voss et al., 2010, Zhang et al., 2014), and links between DMN connectivity and cognitive aging (Andrews-Hanna et al., 2007, Sambataro et al., 2010) corroborate the notion of network-level vulnerability to advancing age (Mowinckel et al., 2012, Salami et al., 2012).

In line with the selective vulnerability of brain networks, age-related cognitive decline is not uniform across domains (Glisky, 2007). Whereas crystallized functions including vocabulary and general knowledge remain relatively unaffected, a commonly reported impairment relates to deficient selective attention (Quigley and Müller, 2014) which manifests as reduced ability to selectively attend to and ignore information based on relevance (Kensinger and Corkin, 2009, Murray and Kensinger, 2012). Whereas resting-state fMRI has provided an opportunity to test specific hypotheses regarding the underlying brain connectivity alterations, the selective cognitive vulnerability suggests that attention-demanding tasks may comprise a more sensitive and informative context for the study of age-related brain network alterations. Thus, in order to characterize the brain functional connectomic signature of cognitive aging in different cognitive contexts, we compare fMRI brain connectivity indices obtained during rest and during two load levels of a continuous multiple object tracking (MOT) task (Pylyshyn and Storm, 1988) between young and older healthy adults.

MOT requires the subject to attend to multiple target items as they move among distractor items, which is a basic feature of sustained multifocal visual attention (Cavanagh and Alvarez, 2005). Alnæs et al. (2015) reported widespread connectivity modulations during MOT in young adults, and machine learning revealed high classification accuracy when discriminating between task and rest, demonstrating successful decoding of cognitive load using FC features. Here, we utilize a similar decoding approach to describe the neuronal characteristics of cognitive aging by comparing two groups of healthy adults at both ends of the adult lifespan. We hypothesized that (1) multivariate classification would yield robust discrimination between resting state and data collected during MOT as well as between younger and older adults, reflecting age and cognitive context related changes in brain network configuration. Higher task classification accuracy in young compared to older adults would support the notion of cognitive and neuronal dedifferentiation in aging. Additionally, we anticipated stronger group differences with higher load, resulting in improved group classification with increasing load. On edge-level, we hypothesized (2) that the most discriminative edges for rest and task are found in connections between attention networks and somatosensory and motor nodes (Alnæs et al., 2015, Tomasi et al., 2014). With regards to age-related edgewise effects in response to attentional demand and keeping with the theory of dedifferentiation, we hypothesized a system-level loss in network segregation, reflected in decreased FC within modular networks and in particular within the DMN as well as increased connectivity between large-scale brain networks, in particular during continuous tracking, which would support a relative inability to selectively inhibit competing neural processes during task engagement.

Finally, since any effects on edgewise functional connectivity may be partly driven by nodal increases or decreases in activity, we probe temporal variability on node level by computing the standard deviation of signal amplitude (SDSA) (Garrett et al., 2010). Brain signal variability has been shown to reveal distinct patterns not captured by mean-based analyses (Garrett et al., 2011), and increased signal variability is thought to be indicative of a more sophisticated and complex neural machinery (McIntosh et al., 2008), offering greater dynamic range and facilitating the brain's ability to explore different network states. When applied to aging studies, reduced SDSA has proved to be a better predictor of advancing age and poorer task performance than the mean measure (Grady and Garrett, 2014). Additionally, in a study investigating age-related variability changes in response to cognitive demand, Garrett et al. (2012) observed a broad increase in SDSA on task compared to fixation for both age groups, the magnitude and spatial extent of this variability increase was comparatively reduced for the older group. Based on this, we hypothesize (3) task related modulations of signal variability that is network selective, reflected in increased and decreased SDSA in task-positive and task-negative/irrelevant networks, respectively. Further, reflecting reduced dynamic repertoire in response to cognitive state-transition, we anticipate group differences in signal variability manifested as diminished task-related SDSA changes for older compared to younger adults.

Section snippets

Sample

We recruited 26 young and 26 older adults through a newspaper ad and social media. The sample is overlapping with a previous publication (Dørum et al., 2016) reporting results from the blocked MOT runs (see below). All subjects provided an informed consent and underwent neuropsychological screening (details below). Participants reported normal or corrected-to-normal vision. Exclusion criteria included estimated IQ<70, previous history of alcohol- and substance abuse, history of neurologic or

Neuropsychology and demographics

Table 1 summarizes key demographics and WASI and MOT performance. Briefly, the younger group performed significantly better on the matrix reasoning subtest of the WASI. Repeated measures ANOVA revealed significant main effects of load (F=17.58, p<.001) and group (F=4.11, p=.049) on MOT accuracy, indicating higher accuracy during L1 compared to L2 and for young compared to old, but no group by load interaction (F=.35, p=.56). No other significant differences were found.

Classification analysis

Fig. 1 shows the confusion

Discussion

The heterogeneity in cognitive aging across cognitive domains is mirrored in a differential vulnerability of brain network connectivity, which has been proposed as a candidate imaging biomarker for age-related cognitive decline and risk and progression of dementia. Since previous studies have primarily utilized resting-state fMRI, we assessed the generalizability of the predictive FC patterns across the resting-state and constrained contexts characterized by higher cognitive load. In line with

Conclusion

Our main findings demonstrate that FC during a constrained continuous tracking task strongly increases sensitivity to age compared to the conventional resting-state paradigm, and secondly, that data obtained during rest and task is more distinguishable in young compared to older individuals, providing further neuronal support of the conjecture of cognitive and neuronal dedifferentiation in aging. Complementary connectivity analyses on node and edge level, as well as nodal signal variability

Acknowledgements

This work was funded by the Research Council of Norway (204966/F20), the South-Eastern Norway Regional Health Authority (2013054, 2014097, 2015073), and the Norwegian Extra Foundation for Health and Rehabilitation (2015/FO5146).

References (83)

  • D.N. Greve et al.

    Accurate and robust brain image alignment using boundary-based registration

    NeuroImage

    (2009)
  • A. Hafkemeijer et al.

    Imaging the default mode network in aging and dementia

    Biochim. Biophys. Acta (BBA)-Mol. Basis Dis.

    (2012)
  • M. Jenkinson et al.

    A global optimisation method for robust affine registration of brain images

    Med. Image Anal.

    (2001)
  • T. Kaufmann et al.

    The brain functional connectome is robustly altered by lack of sleep

    Neuroimage

    (2016)
  • R.E., Jr Kelly et al.

    Visual inspection of independent components: defining a procedure for artifact removal from fMRI data

    J Neurosci. Methods

    (2010)
  • O. Ledoit et al.

    Improved estimation of the covariance matrix of stock returns with an application to portfolio selection

    Journal of empirical finance

    (2003)
  • A.M. Mowinckel et al.

    Network-specific effects of age and in-scanner subject motion: a resting-state fMRI study of 238 healthy adults

    Neuroimage

    (2012)
  • R.H. Pruim et al.

    Evaluation of ICA-AROMA and alternative strategies for motion artifact removal in resting state fMRI

    Neuroimage

    (2015)
  • P.A. Reuter-Lorenz et al.

    Brain aging: reorganizing discoveries about the aging mind

    Curr. Opin. Neurobiol.

    (2005)
  • G. Salimi-Khorshidi et al.

    Automatic denoising of functional MRI data: combining independent component analysis and hierarchical fusion of classifiers

    NeuroImage

    (2014)
  • F. Sambataro et al.

    Age-related alterations in default mode network: impact on working memory performance

    Neurobiol. Aging

    (2010)
  • S.M. Smith et al.

    Group-PCA for very large fMRI datasets

    Neuroimage

    (2014)
  • S.M. Smith et al.

    Advances in functional and structural MR image analysis and implementation as FSL

    Neuroimage

    (2004)
  • S.M. Smith et al.

    Network modelling methods for FMRI

    NeuroImage

    (2011)
  • S.M. Smith et al.

    Network modelling methods for FMRI

    Neuroimage

    (2011)
  • O. Sporns et al.

    Organization, development and function of complex brain networks

    Trends Cogn. Sci.

    (2004)
  • M.P. Van Den Heuvel et al.

    Exploring the brain network: a review on resting-state fMRI functional connectivity

    Eur. Neuropsychopharmacol.

    (2010)
  • Andersson, J.L., Jenkinson, M., Smith, S., 2007a. Non-linear optimisation. FMRIB technical report TR07JA1. FMRIB...
  • Andersson, J.L., Jenkinson, M., Smith, S., 2007b. Non-linear registration, aka Spatial normalisation FMRIB technical...
  • P.B. Baltes et al.

    Emergence of a powerful connection between sensory and cognitive functions across the adult life span: a new window to the study of cognitive aging?

    Psychol. Aging

    (1997)
  • C.F. Beckmann et al.

    Probabilistic independent component analysis for functional magnetic resonance imaging

    IEEE Trans. Med. Imaging

    (2004)
  • D.H. Brainard

    The psychophysics toolbox

    Spat. Vision.

    (1997)
  • S. Bunge

    Differential maturation of brain signal complexity in the human auditory and visual system

    Dev. Human. Brain

    (2009)
  • Chan, M.Y., Park, D.C., Savalia, N.K., Petersen, S.E., Wig, G.S., 2014. Decreased segregation of brain systems across...
  • Chen, S.A., Wu, C.-Y., Lua, R.-p., Akoshi, M.M., Nakai, T., 2013. Age-related changes in resting-state and...
  • Córdova-Palomera, A., Kaufmann, T., Persson, K., Alnæs, D., Doan, N., Moberget, T., Lund, M., Barca, M., Engvig, A.,...
  • M. D'Esposito et al.

    Alterations in the BOLD fMRI signal with ageing and disease: a challenge for neuroimaging

    Nat. Rev. Neurosci.

    (2003)
  • J. Damoiseaux et al.

    Reduced resting-state brain activity in the “default network” in normal aging

    Cereb. Cortex

    (2008)
  • F. Deligianni et al.

    Relating resting-state fMRI and EEG whole-brain connectomes across frequency bands

    Front Neurosci.

    (2014)
  • X. Di et al.

    Modulatory interactions between the default mode network and task positive networks in resting-state

    PeerJ

    (2014)
  • E.S. Dørum et al.

    Age‐related differences in brain network activation and co‐activation during multiple object tracking

    Brain Behav.

    (2016)
  • Cited by (13)

    • Functional brain network modeling in sub-acute stroke patients and healthy controls during rest and continuous attentive tracking

      2020, Heliyon
      Citation Excerpt :

      Each continuous tracking block lasted 7.5 min and contained 14 trials. Detailed outline of the task is described in our previous study (Dørum et al., 2017). Briefly, 10 identical blue objects were presented on a grey background screen, and after another 0.5 s the objects started moving.

    • Does the child brain rest?: An examination and interpretation of resting cognition in developmental cognitive neuroscience

      2020, NeuroImage
      Citation Excerpt :

      It is an open question as to whether or not variability in cognitive context influences functional connectivity measures. There is evidence in adults to suggest that functional network measurements are largely conserved across cognitive contexts (Finn et al., 2015; Gratton et al., 2018), however there is also evidence that variability in functional network measures between contexts can be important indicators of individual tendencies (Finn et al., 2017) and that changes in functional connectivity across adolescence and adulthood are context-dependent (Dørum et al., 2017; Murty et al., 2018). Dørum et al. (2017), examined samples of younger and older adults and found that functional connectivity during a multiple object tracking task more accurately distinguished the two groups than did functional connectivity measured during resting state.

    • A comprehensive analysis of resting state fMRI measures to classify individual patients with Alzheimer's disease

      2018, NeuroImage
      Citation Excerpt :

      We chose an ICA, because it is a data-driven approach that results in spatially independent components well suited for FC analyses. We used a group ICA and imposed the group components onto each subject (Dørum et al., 2017; Miller et al., 2016), because it is important to strive for the same parcellation in each subject in order to compare connectomes across subjects (Smith et al., 2013). We used the ICA components directly as nodes for the FC analysis.

    View all citing articles on Scopus
    1

    postal address: Oslo University Hospital, PoBox 4956 Nydalen, 0424 OSLO, Norway.

    View full text