Effects of aging on default mode network activity in resting state fMRI: Does the method of analysis matter?
Introduction
Imaging neuronal activity using functional magnetic resonance imaging (fMRI) has evolved to an important diagnostic tool to evaluate brain function and neuronal connectivity. A range of studies has described brain regions with synchronous, low-frequency blood oxygen level-dependent (BOLD) signal changes during rest comprising posterior cingulate/precuneus, medial prefrontal and bilateral lateral parietal cortex. Because this network is typically deactivated during external stimulation, it has been termed the ‘default mode network’ (DMN) (Binder et al., 1999, Shulman et al., 1997). The behavioral function of this network is still unresolved. It has been suggested that the DMN plays a role in attending to environmental stimuli as well as mediating processes such as reviewing past knowledge or preparing future actions. It may also be involved in episodic memory (Greicius et al., 2004).
Interestingly, the DMN regions comprise the typical predilection sites of Alzheimer's disease (Mosconi, 2005), the most frequent cause of dementia in the elderly and the most frequent neurodegenerative disorder in humans. Accordingly, resting state fMRI identified significant disruptions in DMN co-activation in patients with AD (Greicius et al., 2004, Rombouts et al., 2007).
Hence, attempts have been made to apply resting state fMRI as a non-invasive, readily available and radiation exposure free biomarker of incipient AD (Greicius et al., 2004). One important prerequisite for the employment of resting state fMRI as a biomarker of AD is a clear understanding of the role of normal aging on DMN connectivity. Age effects on DMN co-activation need to be considered in respect to the specificity of the detection of AD-related abnormalities. Additionally, with a range of possible methods of analysis for fMRI data available, the ability of each method to detect slight age-related changes is an indicator for the sensitivity of the respective test. This information may hence help to identify the most appropriate way of data analysis for a potential future clinical routine use of fMRI in the early detection and differential diagnosis of dementias.
Currently, limited information is available on the influence of physiologic aging on DMN co-activation. It could be shown that the default mode network regions are only sparsely functionally connected at early school age (Fair et al., 2008). During adolescence and early adulthood, these regions seem to integrate into a cohesive, inter-connected network. There is, however, controversy concerning further changes of DMN co-activation with aging in adults, possibly depending on the method of fMRI data processing (Bluhm et al., 2008, Greicius et al., 2004). Either correlations of signal time courses between different brain areas using region of interest analyses (ROI) (Fox et al., 2005, Fransson, 2005) or data-driven extraction of DMN co-activation by independent component analyses (Esposito et al., 2006, Greicius et al., 2004) was used to determine DMN connectivity from fMRI data. Most studies only focused on one of these methods. Substantially different properties of both approaches, however, raise the question, which approach is more sensitive towards age-related changes in DMN connectivity. Time course correlation approaches should, in principal, be more sensitive to true differences in the correlation between specific regions but are limited by a potential variation in the localization of these regions across subjects. Independent component analysis, on the other hand, is likely to be more comprehensive in detecting variation across a well-defined network. However, it may be less sensitive to inter-individual variation in the composition of such networks and may be more likely to produce errors at the group level if a network is presented across multiple components in some subjects. Based on these hypotheses, both approaches will most likely show different sensitivity profiles for the detection of small changes of DMN activity. The detection of such small changes of DMN activity, however, is one of the most important prerequisites for early detection of neurodegenerative dementias.
Therefore, in the present study, we aimed (1) to characterize the effects of normal aging on resting state DMN co-activation as assessed with fMRI by further extending the results of Fair et al. (Fair et al., 2008) to elderly healthy controls and (2) to determine which method, ROI-based correlation analysis or data-driven ICA, is more sensitive to age effects. This information is a prerequisite to better understand the effects of aging on DMN.
Section snippets
Subjects
We prospectively studied two groups of subjects, one comprised of 17 young (7 male, 10 female; mean age ± standard deviation (SD), 27.1 ± 3.0 years; age range, 21.4–32.3 years) and one comprised of 21 older (10 male, 11 female; mean age ± SD, 68.6 ± 7.3 years; age range, 56.4–83.0 years, n = 4 in the 5th decade, n = 7 in 6th decade, n = 9 in the 7th decade and n = 1 in the 80th decade) healthy subjects. Subjects' consent was obtained according to the Declaration of Helsinki. The study was approved by the
Results
All subjects tolerated scanning well. As expected, mean age of the old cohort differed significantly from that of the young cohort (P < .001). Mean ± SD duration of education was 15.0 ± 2.7 years for the young and 13.1 ± 3.8 years for the old subjects with the difference not being significant (Student's t-test P = .103).
Discussion
Our results provide important insights in effects of healthy aging on DMN co-activity as assessed with resting state fMRI. Recently, Fair et al. (2008) could demonstrate that the default mode network is only sparsely functionally connected at early school age and over development integrates to a cohesive, inter-connected network as shown in young adults up to 31 years. Our study further expands these results to older subjects using a similar graph theory approach and also determines differences
Conclusion
Normal aging processes significantly affect effect sizes (an indirect measure of DMN co-activation magnitude) of resting state DMN co-activation, a phenomenon that should be considered when analyzing data with independent component analyses. In contrast, age had no effect on DMN inter-connectivity as determined using signal time course analyses. This might implicate lower sensitivity of inter-connectivity measures to detect subtle DMN changes as those resulting from healthy aging or indicate
Acknowledgments
Funding was obtained through the Science Foundation Ireland (SFI) Stokes Programme to A.L.B. as well as through the SFI investigator neuroimaging programme grant 08/IN.1/B1846 to H.H.
References (52)
- et al.
Disruption of large-scale brain systems in advanced aging
Neuron
(2007) - et al.
Tensorial extensions of independent component analysis for multisubject fMRI analysis
NeuroImage
(2005) - et al.
Separating respiratory-variation-related fluctuations from neuronal-activity-related fluctuations in fMRI
NeuroImage
(2006) - et al.
fMRI resting state networks define distinct modes of long-distance interactions in the human brain
NeuroImage
(2006) The hippocampus and declarative memory: cognitive mechanisms and neural codes
Behav. Brain Res.
(2001)- et al.
Independent component analysis of fMRI group studies by self-organizing clustering
NeuroImage
(2005) - et al.
Independent component model of the default-mode brain function: assessing the impact of active thinking
Brain Res. Bull.
(2006) - et al.
Changes in glucose metabolism due to aging and gender-related differences in the healthy human brain
Psychiatry Res.
(2008) - et al.
Large-amplitude, spatially correlated fluctuations in BOLD fMRI signals during extended rest and early sleep stages
Magn. Reson. Imaging
(2006) - et al.
A voxel-based morphometric study of ageing in 465 normal adult human brains
NeuroImage
(2001)