Working memory in Aging. Graph Theory for MEG-DTI study
-
1
Center for Biomedical Technology, Neuroimaging, Spain
-
2
Center for Biomedical Technology, Cognitive and Computational Neuroscience, Spain
-
3
Center for Biomedical Technology, Biological Networks, Spain
-
4
Rey Juan Carlos University, Complex Systems Group, Spain
INTRODUCTION.
The study of changes in phase synchronization on MEG recordings provides interesting clues of how memory-related brain connectivity patterns arise and how they are sustained during the presentation of interfering information. In this work we compute graph metrics from MEG data describing precisely how the brain networks are impaired during an active interference, and we correlate those results with the fractional anisotropy images from the same subjects, revealing which structures exert high influence over the maintenance of information. This combination of information constitutes a novel approach which allows to analyze both structural and functional information.
MATERIALS AND METHODS.
MEG - task: An adapted delayed match-to-sample audiovisual task divided into three stages (Fig. 1): encoding, maintenance with active interference and recognition. The study is focused on the interference stage, specifically on the 500ms prior to the interference (pre) and on the 500ms period after the 3s of interference (post) for both hits and errors.
Subjects: 8 elderly women with ages 65.56 ± 6.06 years. MEG recordings were obtained in a 148-channel (magnetometer) MAGNES 2500 WH, 4D Neuroimaging. Diffusion weighted images were obtained in a 1.5T Signa Excite MR Scanner (GE Healthcare) with b=0;900s/mm2 and 25 gradient directions.
Synchronization Analysis: A phase synchronization index (Phase Locking Value - PLV) was computed for each pre and post interference 500ms-period in 5 frequency bands.
Graph Analysis: A weighted graph for each PLV matrix was created. Network Strength (S), characteristic path length (L) and clustering (C) coefficients were obtained for each of the networks and normalized over randomized networks.
Statistical Analysis: All scans were processed with FSL 3.1.8 for motion correction and eddy currents artifacts, and a fractional anisotropy (FA) map was obtained for each subject. SPM8 was used to compute a FA generic template [1] and to perform statistical analysis. A general linear model was set up for each graph metric per frequency band, adding age as a confound regressor.
RESULTS.
Graph Analysis over MEG PLVs networks: When there is an error, S, in the post period, is higher than in any other conditions (Fig. 2a). This fact indicates a higher synchronized activity in the post- errors, despite that it seems that the functional network does not succeed in finding the information it is looking for. Furthermore, in the post-interference period preceding a hit response, the strength value decreased slightly. All together indicates that the higher the strength of the network the higher possibility to make an error during the recognition stage. C increases in most of the frequency bands between both periods when there is an error (Fig. 2b). However in the presence of hit responses the C values decrease, although differences between these two periods of time are close to zero. Additionally, the previously reported increases of S lead to a reduction of L (Fig. 2c).
Normalized graph metrics (Cnorm and Lnorm) offer a different point of view when analyzing brain networks, since the influence of an increase/decrease of the network strength is ruled out by the normalization. On the contrary, normalized parameters reveal changes in the network organization, being close to one when the networks reorganize in a more random configuration.
In the maintenance period preceding a hit response, Cnorm does not change much its organization between pre- and post-interference periods, but in the post-interference period preceding an error response, a decrease takes place (Fig. 3a), indicating that the network became more random (and more synchronized as Fig. 2a). This reveals a decrease of local connections. Lnorm behaves exactly the same way (Fig. 3b), showing a more random configuration in the case of post- errors.
Correlations over FA images: We perform statistics over the FA images including the Lnorm as variable of interest. Multiple regression statistic maps were threshold with 0.005 p-value (uncorrected) under a cluster coefficient of 50 voxels. This p-value is not low enough in order to affirm that our correlations are due exclusively to the graph metric, rather than to random dependencies, but due to the reduced number of subjects, we just achieve p-values of 10-5, corresponding to voxels alone in the same regions that have been depicted. We observe and conclude that normalized shortest path Lnorm, at alpha, beta and gamma frequency bands, correlates positively with the integrity of the white matter at the frontal lobe regions (Fig. 4), in the context of hit responses and at the pre-interference period of time.
CONCLUSIONS.
This study provides a series of evidences indicating how we do forget when we get elders. The neurophysiological mechanisms underlying forgetting are represented by an increase of the strength of the network combined with a random structure as indicated by the C and L normalized parameters. All these phenomena occur in the post-interference period of time preceding an unsuccessful recognition (error response) at the recognition stage. Furthermore, the correlation between the functional and the fractional anisotropy values indicate that the higher the integrity of the white matter at the frontal lobe regions (and medial temporal lobe in pre-interference), in both the pre- and the post-interference periods of maintenance, the more the efficient functional network organization based on the L parameter. All these, is leading us to, at least, three provocative ideas: 1) the increase of the strength is related with a random architecture and a high cost of energy indicating an inefficient network architecture to prevent for memory interference; 2) the importance of some of the prefrontal white matter regions in the control of interference; 3) the possible impairment of these white matter tracts will produce a dramatic consequence in the functional architecture of the whole network because these regions are the paths used to transmit information from the frontal lobe to several regions of the limbic and posterior cortex. It is necessary to compare all these parameters with those of young subjects and having additional control conditions.
References
[1] Focke N. et al. Voxel.based diffusion tensor imaging in patients with mesial temporal lobe epilepsy and hippocampal sclerosis. Neuroimage 40, 728-737. 2008.
Keywords:
Diffusion weighted image,
graph theory,
Magnetoencephalography,
working memory
Conference:
XI International Conference on Cognitive Neuroscience (ICON XI), Palma, Mallorca, Spain, 25 Sep - 29 Sep, 2011.
Presentation Type:
Poster Presentation
Topic:
Poster Sessions: Cognitive Aging
Citation:
Pineda Pardo
J,
Niso
J,
Solesio
E,
Hernandez-Tamames
J,
Buldú
JM,
Maestú
F and
Del Pozo
F
(2011). Working memory in Aging. Graph Theory for MEG-DTI study.
Conference Abstract:
XI International Conference on Cognitive Neuroscience (ICON XI).
doi: 10.3389/conf.fnhum.2011.207.00061
Copyright:
The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers.
They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.
The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.
Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.
For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.
Received:
15 Nov 2011;
Published Online:
25 Nov 2011.
*
Correspondence:
Mr. Jose Angel Pineda Pardo, Center for Biomedical Technology, Neuroimaging, Boadilla del Monte, Madrid, 28223, Spain, joseangel.pineda@ctb.upm.es