Quantitative Three-Dimensional Characterization of Facial Dysmorphism in Self-Identified African-American Individuals With Prenatal Alcohol Exposure.

Alzheimer's disease (AD) is an age‐related neurodegenerative disorder pathologically characterized by beta‐amyloid (A) plaques and neurofibrillary tangles, accompanied by progressive cognitive decline. AD neuropathology, particularly A, develops years to decades prior to the onset of clinical symptoms (Jack et al. ; Sperling et al. ), and as a consequence, alterations may occur to structural and functional neural systems that eventually lead to pathological cognitive decline. Though much knowledge has been garnered regarding the mechanisms underlying AD, the etiology and biological progression of AD remain elusive. With the number of individuals affected by AD projected to more than double by 2050 (Hebert et al. ), a better understanding of the mechanisms underlying AD is needed to develop more targeted treatments and to identify biological markers to aid early intervention, when treatments may be most effective. AD is proposed to be a disconnection syndrome (Delbeuck et al. ), where the precise spatiotemporal functional coherence of brain regions within and between neural networks is disrupted potentially as a downstream consequence of neuropathology, which underpins cognitive decline. One technique to better characterize network disruptions across the AD continuum is graph theoretical analysis of resting state functional magnetic resonance imaging (rsfMRI) data (Dai and He ). Graph theory models the brain and its functional connections as a network and examines the topology that underlies higher‐order information processing and integration. A graph is a mathematical representation of a brain network and is constructed as a series of nodes and edges (Rubinov and Sporns ). Nodes are comprised of a set of brain regions while edges represent the functional connection, or correlation, between the timeseries of each node. From these graphs, two global properties of the network can be explored including segregation and integration. Network segregation represents the ability for specialized processing to occur within clusters of highly connected regions of the network (Rubinov and Sporns ). In functional networks, segregation properties describe the number of submodules present within the network exhibiting high intra‐module connectivity. Network integration captures the ability of the network to rapidly transfer and combine information across specialized modules and distributed nodes (Rubinov and Sporns ). Importantly, functional networks are shown to exhibit an optimal balance between segregation and integration, such that the network is structured to support subnetwork organization to perform specialized processing but also exhibits strategically placed links to facilitate efficient information integration across submodules. This network organizational efficiency can be quantified through small‐worldness (Humphries and Gurney ). Individual differences in global network organization are related to cognitive performance in samples of healthy adults (Bertolero et al. ; Cohen and D'Esposito ; Stumme et al. ) and alterations to such properties have functional significance in patient samples, with disrupted network segregation, for example, related to greater AD clinical impairment (Brier et al. ; Dai et al. ). Thus, understanding how the global topology of functional networks is affected by age and disease will provide critical information regarding macroscale mechanisms underlying pathological age‐related cognitive impairment. AD‐related pathological aging has been associated with differences in both global network properties (segregation, integration) (Brier et al. ; Khazaee et al. ; Li et al. ; Si et al. ; Supekar et al. ; Wang et al. ) and local nodal attributes (Behfar et al. ; Luo et al. ; Subramanian et al. ), features which can be used with high fidelity to distinguish cognitively normal (CN), mild cognitive impairment (MCI), and AD groups (de Vos et al. ; Hojjati et al. , ; Khazaee et al. ; Liu et al. ; Xu et al. ). While significant differences in association with AD and AD pathologies have been reported in these network properties, the directionality of these differences vary across studies, which impedes efforts to draw conclusions about the mechanisms underlying network dysfunction and their functional consequences across the AD spectrum. For example, studies investigating clustering coefficient, a measure of network segregation, have reported higher clustering coefficient in MCI or AD than CN (Liu et al. ; Zhao et al. ), lower clustering coefficient in MCI or AD than CN (Brier et al. ; Dai and He ; Dai et al. ; Li et al. ; Si et al. ; Xiang et al. ; Xue et al. ), or no diagnostic group differences (Khazaee et al. ; Liu et al. ; Sanz‐Arigita et al. ; Zhang et al. ). Similar discrepancies are reported with characteristic path length (a measure of network integration) (Brier et al. ; Dai et al. ; Khazaee et al. ; Li et al. ; Si et al. ; Supekar et al. ; Xiang et al. ; Zhang et al. ; Zhao et al. ), and sm