Effect of education on functional network edge efficiency in Alzheimer's disease

Abstract

We investigated the effect of education on the edge efficiency in resting state functional networks (RSFNs) in amnestic mild cognitive impairment (aMCI) and Alzheimer's disease dementia (ADD). We collected the data of 57 early aMCI, 141 late aMCI, 173 mild ADD, and 39 moderate-to-severe ADD patients. We used years of education as a proxy for cognitive reserve. We measured edge efficiency for each edge in RSFNs, and performed simple slope analyses to discover their associations with education level among the four groups. In the late aMCI, a sub-network that had hub nodes in the right middle frontal gyrus and the right posterior cingulate gyrus, showed a positive association between RSFN edge efficiency and education (threshold = 2.5, p = 0.0478). There was no negative effect of education on the RSFN edge efficiency. In the early aMCI, mild ADD, and moderate-to-severe ADD, there were no sub-networks showing positive or negative correlation between education and RSFN edge efficiency. There was a positive effect of higher education on RSFN edge efficiency in the late aMCI, but not in the early aMCI or ADD. This indicates that in late aMCI, those who have higher education level have greater ability to resist collapsed functional network.

Figure 1

(A) In the late aMCI group, a sub-network showed a positive correlation between years of education and RSFN edge efficiency. The red circles represent the hub regions, which are located in the orbital part of the right middle frontal gyrus (MFGorb), and the right posterior cingulate gyrus (PCC). (B) Connectogram showing the sub-network. The thickness and colors represents the t-statistics computed from simple slope analyses. Hub regions are presented in red color. (C) Correlation between education and mean RSFN edge efficiency of the sub-network in the late aMCI group. aMCI amnestic mild cognitive impairment.

Figure 2

Schematic overview of the proposed method. (A) The concept of edge efficiency is depicted with an example. In this example, edges in network A and network B are same, except that the edge (i, c) does not exist in the network B. Although the exchanges of information through the edge (i, j) (red line and magenta arrow) are similar in both networks, different network topology results in different edge efficiency values for the edge (i, j) in the two networks. When the edge (i, j) is removed, information communicating through other paths alternative to the edge (i, j) is different in network A and B. (B) Network-based statistical analysis to identify sub-networks whose edge efficiency values are significantly associated with education years. The analysis was performed using cluster-based statistics. Note that it detects significant sub-networks, and it cannot extract any single significant edge such as the edge (i, c) in network A.