An exponential random graph modeling approach to creating group-based representative whole-brain connectivity networks.

Group-based brain connectivity networks have great appeal for researchers interested in gaining further insight into complex brain function and how it changes across different mental states and disease conditions. Accurately constructing these networks presents a daunting challenge given the difficulties associated with accounting for inter-subject topological variability. Viable approaches to this task must engender networks that capture the constitutive topological properties of the group of subjects' networks that it is aiming to represent. The conventional approach has been to use a mean or median correlation network (Achard et al., 2006; Song et al., 2009; Zuo et al., 2011) to embody a group of networks. However, the degree to which their topological properties conform with those of the groups that they are purported to represent has yet to be explored. Here we investigate the performance of these mean and median correlation networks. We also propose an alternative approach based on an exponential random graph modeling framework and compare its performance to that of the aforementioned conventional approach. Simpson et al. (2011) illustrated the utility of exponential random graph models (ERGMs) for creating brain networks that capture the topological characteristics of a single subject's brain network. However, their advantageousness in the context of producing a brain network that "represents" a group of brain networks has yet to be examined. Here we show that our proposed ERGM approach outperforms the conventional mean and median correlation based approaches and provides an accurate and flexible method for constructing group-based representative brain networks.

Moderate-Heavy Alcohol Consumption Lifestyle in Older Adults Is Associated with Altered Central Executive Network Community Structure during Cognitive Task.

Older adults today consume more alcohol than previous generations, the majority being social drinkers. The effects of heavy alcohol use on brain functioning closely resemble age-related changes, but it is not known if moderate-heavy alcohol consumption intensifies brain aging. Whether a lifestyle of moderate-heavy alcohol use in older adults increased age-related brain changes was examined. Forty-one older adults (65-80 years) that consumed light (< 2 drinks/week and ≥ 1 drink/month, n = 20) or moderate-heavy (7-21 drinks/week, non-bingers, n = 21) amounts of alcohol were enrolled. Twenty-two young adults (24-35 years) were also enrolled (light, n = 11 and moderate-heavy, n = 11). Functional brain networks based on magnetic resonance imaging data were generated for resting state and during a working memory task. Whole-brain, Central Executive Network (CEN), and Default Mode Network (DMN) connectivity were assessed in light and moderate-heavy alcohol consuming older adults with comparisons to young adults. The older adults had significantly lower whole brain connectivity (global efficiency) and lower regional connectivity (community structure) in the CEN during task and in the DMN at rest. Moderate-heavy older drinkers did not exhibit whole brain connectivity differences compared to the low drinkers. However, decreased CEN connectivity was observed during the task. There were no differences in the DMN connectivity between drinking groups. Taken together, a lifestyle including moderate-heavy alcohol consumption may be associated with further decreases in brain network connectivity within task-related networks in older adults. Further research is required to determine if this decrease is compensatory or an early sign of decline.

Long-term moderate alcohol consumption does not exacerbate age-related cognitive decline in healthy, community-dwelling older adults.

Recent census data has found that roughly 40% of adults 65 years and older not only consume alcohol but also drink more of it than previous generations. Older drinkers are more vulnerable than younger counterparts to the psychoactive effects of alcohol due to natural biological changes that occur with aging. This study was specifically designed to measure the effect of long-term moderate alcohol consumption on cognitive health in older adult drinkers. An extensive battery of validated tests commonly used in aging and substance use literature was used to measure performance in specific cognitive domains, including working memory and attention. An age (young, old) (*) alcohol consumption (light, moderate) factorial study design was used to evaluate the main effects of age and alcohol consumption on cognitive performance. The focus of the study was then limited to light and moderate older drinkers, and whether or not long-term moderate alcohol consumption exacerbated age-related cognitive decline. No evidence was found to support the idea that long-term moderate alcohol consumption in older adults exacerbates age-related cognitive decline. Findings were specific to healthy community dwelling social drinkers in older age and they should not be generalized to individuals with other consumption patterns, like heavy drinkers, binge drinkers or ex-drinkers.

Defining nodes in complex brain networks.

Network science holds great promise for expanding our understanding of the human brain in health, disease, development, and aging. Network analyses are quickly becoming the method of choice for analyzing functional MRI data. However, many technical issues have yet to be confronted in order to optimize results. One particular issue that remains controversial in functional brain network analyses is the definition of a network node. In functional brain networks a node represents some predefined collection of brain tissue, and an edge measures the functional connectivity between pairs of nodes. The characteristics of a node, chosen by the researcher, vary considerably in the literature. This manuscript reviews the current state of the art based on published manuscripts and highlights the strengths and weaknesses of three main methods for defining nodes. Voxel-wise networks are constructed by assigning a node to each, equally sized brain area (voxel). The fMRI time-series recorded from each voxel is then used to create the functional network. Anatomical methods utilize atlases to define the nodes based on brain structure. The fMRI time-series from all voxels within the anatomical area are averaged and subsequently used to generate the network. Functional activation methods rely on data from traditional fMRI activation studies, often from databases, to identify network nodes. Such methods identify the peaks or centers of mass from activation maps to determine the location of the nodes. Small (~10-20 millimeter diameter) spheres located at the coordinates of the activation foci are then applied to the data being used in the network analysis. The fMRI time-series from all voxels in the sphere are then averaged, and the resultant time series is used to generate the network. We attempt to clarify the discussion and move the study of complex brain networks forward. While the "correct" method to be used remains an open, possibly unsolvable question that deserves extensive debate and research, we argue that the best method available at the current time is the voxel-wise method.

Consistency of network modules in resting-state FMRI connectome data.

At rest, spontaneous brain activity measured by fMRI is summarized by a number of distinct resting state networks (RSNs) following similar temporal time courses. Such networks have been consistently identified across subjects using spatial ICA (independent component analysis). Moreover, graph theory-based network analyses have also been applied to resting-state fMRI data, identifying similar RSNs, although typically at a coarser spatial resolution. In this work, we examined resting-state fMRI networks from 194 subjects at a voxel-level resolution, and examined the consistency of RSNs across subjects using a metric called scaled inclusivity (SI), which summarizes consistency of modular partitions across networks. Our SI analyses indicated that some RSNs are robust across subjects, comparable to the corresponding RSNs identified by ICA. We also found that some commonly reported RSNs are less consistent across subjects. This is the first direct comparison of RSNs between ICAs and graph-based network analyses at a comparable resolution.