آسیب های مغزی خفیف: خصوصیات مدل گراف شبکه مغزی برای حافظه اپیزودیک
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|33651||2011||8 صفحه PDF||سفارش دهید||7925 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : International Journal of Psychophysiology, Volume 79, Issue 2, February 2011, Pages 89–96
Episodic memory is among the cognitive functions that can be affected in the acute phase following mild traumatic brain injury (MTBI). The present study used EEG recordings to evaluate global synchronization and network organization of rhythmic activity during the encoding and recognition phases of an episodic memory task varying in stimulus type (kaleidoscope images, pictures, words, and pseudowords). Synchronization of oscillatory activity was assessed using a linear and nonlinear connectivity estimator and network analyses were performed using algorithms derived from graph theory. Twenty five MTBI patients (tested within days post-injury) and healthy volunteers were closely matched on demographic variables, verbal ability, psychological status variables, as well as on overall task performance. Patients demonstrated sub-optimal network organization, as reflected by changes in graph parameters in the theta and alpha bands during both encoding and recognition. There were no group differences in spectral energy during task performance or on network parameters during a control condition (rest). Evidence of less optimally organized functional networks during memory tasks was more prominent for pictorial than for verbal stimuli.
Mild Traumatic Brain Injury (MTBI) is the most common cause of brain insult. According to the Mild Traumatic Brain Injury Committee of the Head Injury Interdisciplinary Special Interest Group (American Congress of Rehabilitation Medicine; Kay et al., 1993), a patient with traumatically induced disturbance of brain functions with loss of consciousness for 30 min or less, posttraumatic amnesia not greater than 24 h and an initial Glasgow Coma Scale score of 13–15, is considered to have mild traumatic brain injury. MTBI may affect many cognitive functions, including attention (Alexander, 1995, Binder et al., 1997, De Monte et al., 2006, Vanderploeg et al., 2005 and Voller et al., 1999). Memory is also affected and apart from the abundant reports concerning working memory deficits (Levin et al., 1987, Ruff et al., 1989 and Vanderploeg et al., 2005), there are also reports of deficits in episodic memory, including verbal learning (De Monte et al., 2006, Ruff et al., 1989 and Voller et al., 1999) and visual memory (Levin et al., 1987, Raskin, 2000 and Ruff et al., 1989). For instance, Ruff et al. (1989) followed 32 MTBI patients between 7 days and three months following injury. Whereas patients as a group scored significantly lower than a control group at the initial assessment on the Mattis–Konver Verbal Learning and Memory Test and the Benton Visual Retention Test, these differences were minimized at follow-up. There are also reports for a general reduction in processing speed, especially in the acute phase following a MTBI, which is more prominent during difficult, cognitively demanding tasks (e.g. Gentilini et al., 1989, Gronwal, 1989 and Landre et al., 2006). These deficits, however, are often very mild and not clinically significant at the individual patient level. In one of the few studies of material-specific episodic memory difficulties in the acute phase following MTBI, Tsirka et al. (2010) examined 26 MTBI patients and 26 age-, education, and IQ-matched controls on a series of neuropsychological tests evaluating primary and secondary memory, executive functions, and verbal fluency. Participants were also tested on experimental episodic memory tasks involving words, pseudowords, pictures of common objects and abstract kaleidoscopic images. In addition to reduced performance on standardized episodic memory and executive measures, MTBI patients showed evidence of increased response bias during recognition. The presumed difficulties among MTBI patients were more evident on tasks involving pictorial and, especially complex, abstract, stimuli, such as pictures and kaleidoscope images. The present study explores potential correlates of the relatively subtle, task-specific changes in performance reported by Tsirka et al. (2010) on EEG measures obtained during performance of the same tasks. Several investigations have applied electrophysiological methods (EEG, MEG, and ERPs) to study brain function in MTBI (e.g. Ford and Khalil, 1996, Gaetz et al., 2000, Gaetz, 2002, Lewine et al., 1999, Nuwer et al., 2005, Pointinger et al., 2002, Gaetz and Bernstein, 2001, Nuwer, 1997, Wallace et al., 2001 and Thatcher et al., 1989), although EEG studies of memory task-related changes in MTBI are scarce. In normative studies, episodic memory has been studied extensively with the majority of studies focusing on EEG oscillations in the alpha and theta bands (e.g., Doppelmayr et al., 1998, Klimesch et al., 2001a, Klimesch et al., 2008 and Sauseng et al., 2008). Recently developed methods of network analysis applied to bioelectrical signals have emerged as promising tools for the evaluation of brain function in real time. Instead of focusing on regional task-specific changes in EEG parameters, like power measures, these methods rely on the premise that the brain is a complex network of interconnected regions, each hosting one or more neurophysiological processes, which are engaged in unison in order to accomplish a particular psychological function. One of these measures, synchronization likelihood (SL), in contrast to traditional coherence measures, is sensitive to both linear and nonlinear statistical independencies between a time series within a dynamic system (Stam and van Dijk, 2002). SL can then be used to evaluate the properties of hypothetical, underlying neural networks using mathematical models provided by graph theory. A graph is a network model characterized by nodes (vertices) and connections (edges) between nodes. For EEG data, nodes are represented by electrode sites, whereas SL values for each and every pair of electrodes serve as the edges of the graph model. Widely known direct indices of network organization provided by graph-model algorithms are the clustering coefficient (Cp) and characteristic path length (Lp). The former is an index of the connectedness between neighbouring nodes (if A connects to B and C, what is the chance that B and C will be interconnected as well). The latter is the average shortest path connecting any two vertices; the number of edges it contains indicates the length of a path. Lp is an indicator of the overall connectedness (integration) of the graph. These two measures form the basis for computing an index of “small-world” network (SWN) organization ( Humphries and Gurney, 2008). A SWN is an ideal network model consisting of several local and a few random distant connections and is characterized by a high clustering coefficient and a short path length. Optimal SWN organization has been found to characterize anatomical, hemodynamic, and electrophysiological data obtained from healthy individuals ( Bullmore and Sporns, 2009, Micheloyannis et al., 2006b and Stam and Reijneveld, 2007). Optimally functioning network models, such as those computed from multielectrode, surface EEG recordings are believed to reflect optimally organized, underlying brain circuits engaged during performance of a particular function—in the present study, memory encoding and recognition. Whether this is indeed the case is an empirical question which should be reassessed in each study. An indirect approach to establishing the validity of this assumption is to demonstrate significant associations between network model parameters during performance of memory tasks and memory ability measures obtained independently. Secondary, episodic memory functions such as those tapped by the task used in the present study are suitable for this purpose, as they require engagement of a tightly interconnected set of brain areas (for a recent review see Dash et al., 2006). In contrast, brain dysfunction is expected to alter network structure, generally in the direction of less optimal SWN organization, indicating more “random” networks. Binary, unweighted graph parameters are more commonly used in relevant EEG studies, with weighted graphs emerging more recently as a method of choice, as they specify indices of the strength of modelled connections ( Alstott et al., 2009, Bassett and Bullmore, 2009, Micheloyannis et al., 2006a, Ponten et al., 2009, Rubinov et al., 2009 and Stam et al., 2009). Here we employ a variety of EEG-derived measures, including power in individually-determined frequency bands (according to Klimesch et al., 1994, Klimesch et al., 1996 and Klimesch, 1999), Synchronization Likelihood, clustering, path length, and SWN (using both weighted and unweighted calculation methods). The main goal of the study was to assess global cortical connectivity and network organization in the acute phase of MTBI in comparison to age-, education-, gender-, and IQ-matched, neurologically healthy control participants. Estimates of network organization in the same subjects were obtained at rest and during the encoding and recognition phases of the same episodic memory tasks reported by Tsirka et al. (2010). The MTBI patients were expected to differ from control participants over a wide range of frequencies including theta and three, individually-determined alpha sub-bands. The role of stimulus type was also examined. It was expected that, during encoding and recognition of pictorial stimuli, especially of the more abstract kaleidoscope images, findings will be indicative of deviating from the optimal small-world configuration i.e. of less optimal network organization in MTBI patients.