Chapter 8 – Network Modeling of Epilepsy Using Structural and Functional MRI




Chapter 8 Network Modeling of Epilepsy Using Structural and Functional MRI


Lorenzo Caciagli , Boris C. Bernhardt , Andrea Bernasconi and Neda Bernasconi



8.1 Introduction


Since the first classification attempts, major advancements in our understanding of human epileptogenesis have posed notable challenges to the conventional models of “focal” and “generalized” epilepsies. A wealth of observations from experimental paradigms, animal models, and studies in humans indicates that specific cortical and subcortical networks are involved in the generation of focal and generalized seizures, suggesting that epilepsies may be conceptualized as disorders of neural networks.1, 2



8.2 Epilepsy as a Network Disorder


In temporal lobe epilepsy (TLE), which has been long considered the prototypical focal epilepsy syndrome, studies employing intracranial electroencephalography (EEG) demonstrated that seizure activity involves a widespread set of regions that consists of mesiotemporal, neocortical, as well as subcortical structures.1 Advancements in functional and structural magnetic resonance imaging (MRI) techniques have further revealed complex connectional derangements.3, 4 Initial evidence of widespread structural and functional reconfigurations is also emerging for epilepsies secondary to cortical malformations, particularly focal cortical dysplasia.4 Moreover, several studies suggest that abnormalities outside the lesional boundaries may negatively impact the outcome of epilepsy surgery, which is still suboptimal in up to 40% of candidates despite rigorous selection.58 Collectively, these findings have prompted a major conceptual shift from the conventional interpretation of focal epilepsies, and emphasize the importance of a network approach to adequately capture the neurobiology of these disorders.


In idiopathic generalized epilepsies (IGE), it is widely accepted that an aberrant thalamocortical interaction plays a major role in the generation of spike-wave discharges.9 Several lines of evidence from experimental and animal models indicate that while some cortical regions may play a pivotal role in seizure generation, others would be relatively uninvolved.10 EEG and magnetoencephalography (MEG) analyses focusing on the initiation of human absence seizures demonstrated the predominant involvement of focal cortical areas, mostly represented by mesial frontal and orbitofrontal cortices.1113 In addition, imaging studies reported a rather localized pattern of structural and functional abnormalities in several IGE syndromes.2, 14 Findings indicate that generalized epilepsies might be better understood as arising from the pathological activity of neuronal networks encompassing specific corticosubcortical structures. This is also reflected in the recently formulated concept of system epilepsies, which would view some of the IGE syndromes as the pathologic expression of specific neural systems, normally subserving identifiable functions in the healthy brain.15


Hence, adopting a network perspective seems nowadays compelling to fully capture the complexity of human epilepsies. In this regard, a well-established role is attributed to noninvasive imaging techniques, given their ability to probe connectivity in vivo at multiple scales and with complementary modalities. In the following paragraphs, we discuss the underpinnings of network analysis in epilepsy from an imaging standpoint, focusing on the most widely employed MRI methods. We first summarize the evidence obtained for focal epilepsy syndromes, placing particular emphasis on TLE and extratemporal epilepsy secondary to focal cortical dysplasia (FCD). A subsequent paragraph details findings in patients with IGE. Finally, we outline the relevance of network modeling techniques for clinical decision making.



8.3 Network Modeling Using Functional and Structural MRI


At its simplest, a network can be envisioned as a collection of items with pairwise relationships. This concept can be translated at multiple levels; the brain as a whole can be considered a hierarchically organized network, partitioned into mutually interconnected units responsible for information processing spanning from local circuits to broad functional areas.16 The shift to a network perspective has assumed a particular relevance in epilepsy, since structures that are part of an epileptogenic network are thought to be involved in the generation and expression of seizures, and to the maintenance of the disorder.1


Structural connectivity refers to direct anatomical associations between brain regions.17 The two major MRI techniques employed to map structural networks in humans are diffusion-weighted imaging and covariance analysis (or morphometric correlations) (Figure 8.1). Diffusion imaging provides voxel-wise information about the magnitude and directionality of water diffusion and is utilized to assess the microstructural integrity of the white matter. The use of tractography algorithms further allows reconstructing fiber pathways along plausible diffusion trajectories.18, 19 Despite limitations in signal disambiguation for voxels containing intersecting fibers, results of diffusion imaging analyses show a high degree of consistency and reproducibility and have been cross-validated against tract-tracing studies conducted on animal models.2023 An alternative technique to generate structural network representations of the human brain is MRI covariance analysis. This framework infers networks from interregional correlations of structural markers, such as gray matter volume or cortical thickness. The existence of a network link between two regions is derived from a high correlation of morphological markers between them, while a low correlation speaks against a link. Covariance patterns of structural markers may reflect trophic and/or signaling interactions among distant areas,24 exhibit high correspondence with maturational networks,25 and overlap with networks derived via diffusion imaging and resting-state fMRI.2628 Differences in structural measures also appear to covary within assemblies of brain regions belonging to well characterized functional systems, such as those subserving visual, auditory, motor, language, and other cognitive functions.24, 29, 30 Thus, covariance patterns seem to reflect brain connectivity, and may aid in unveiling interregional pathological influences in the context of brain disorders.





Figure 8.1. MRI methods to model networks in the human brain. The passages leading to derive brain networks through different MRI modalities (diffusion MRI, functional MRI, and structural MRI covariance) are here summarized. Within each modality, data are usually preprocessed and analyzed in parcellated anatomical space. Tractography algorithms permit reconstructing fiber pathways between parcels along plausible diffusion trajectories. Resting-state fMRI and structural MRI covariance rely on cross-correlations between seed and target regions. A connectome-based approach implies the iterative assessment of pairwise associations between parcels, and leads to derive connectivity matrices. An alternative representation of a connectivity matrix is a graph, in which nodes correspond to brain areas and edges represent connections.


Adapted from Bernhardt et al., Epilepsy Behav 2015 with permission.

Functional connectivity refers to statistical associations between spatially distributed neurophysiological time series.31 Functional connectivity measures can be derived from a variety of signal sources, spanning from electrophysiological techniques, such as EEG and MEG, to imaging methods, the most prominent of which is represented by functional MRI (fMRI). In fMRI, changes in blood-oxygenation-level-dependent (BOLD) signal are utilized to infer neuronal activity under a neurovascular coupling model. While the temporal resolution of fMRI is generally low (order of seconds), its spatial resolution falls in the millimeter range and whole-brain coverage is permitted. A versatile tool is represented by resting-state fMRI paradigms, during which subjects lie still in the scanner without performing any tasks (Figure 8.1). Acquiring resting-state fMRI datasets exhibits several advantages compared with task-based paradigms, including the possibility to assess multiple regions simultaneously, as well as reduced demand for patients with impaired ability to engage in tasks.32 Analysis of such sequences has led to the identification of brain networks, which show coherent fluctuations in their intrinsic, spontaneous activity. Resting-state networks display a high degree of reproducibility across subjects, as well as robust correspondence with task-involved systems.33 Hence, resting-state fMRI can be utilized to probe intrinsic functional networks and assess disease-specific connectional derangements.34


In recent years, graph theory has emerged as a unique framework to characterize network organizational properties at a whole-brain level.35 Graph theory formalizes a network as a collection of nodes, corresponding to brain regions, interconnected by pairwise edges, derived from structural and functional connectivity estimates (Figure 8.1). Nodes can often be clustered into modules, which show a dense internal connectivity but a relative segregation from the rest of the network. Centrality-based metrics allow identifying hubs, i.e., nodes with a high degree of connections to the rest of the network and prominent roles in network dynamics. Another set of measures addresses the efficiency of local and global information transmission, such as the clustering coefficient, which quantifies connection density within the local environment subnetwork surrounding a node, and path length, which describes the average number of connections between any pairs of nodes. In healthy individuals, whole-brain topological properties exhibit small world attributes.36 Combining high levels of clustering with overall short paths, a small world architecture represents an efficient arrangement to achieve segregation and integration of information processing while minimizing wiring costs.37 Graph theoretical analysis provides the unique opportunity to assess high-order properties of functional and structural networks, with the possibility to span from rather localized domains to a whole-brain level.



8.4 Evidence for Extensive Network Disruptions in Temporal Lobe Epilepsy


Studies probing the integrity of structural and functional networks in TLE have found evidence of widespread abnormalities affecting temporolimbic circuits as well as several large-scale networks, along with striking reconfigurations of whole-brain organizational properties (Figure 8.2). Morphometric correlation analyses of cortical thickness and gray matter volume revealed decreased structural coordination between mesiotemporal regions and an extensive assembly of neocortical areas, including lateral temporal neocortices38, 39 as well as prefrontal, frontocentral, occipitotemporal, and cingulate cortices.3840 Covariance of thalamic atrophy with cortical thickness of mesiotemporal,41 frontocentral, and lateral temporal neocortices42 has been detected, pointing to a prominent involvement of the thalamus in the pathologic network of TLE. Considering the underlying white matter, abnormal diffusion parameters have been observed for several temporolimbic tracts both ipsilateral and contralateral to the seizure focus, including fornix, cingulum, and uncinate fasciculus.4347 Moreover, reduced fractional anisotropy has been documented for several extratemporal bundles, including the superior and inferior longitudinal fascicles, occipitofrontal fascicle, external and internal capsules, and corpus callosum.44, 45, 4749 Diffusion derangements also encompass specific connections arising from the thalamus, such as thalamo-mesiotemporal fibers, ipsilateral anterior thalamic radiation, and tracts linking ipsilateral thalamus with the precentral gyrus.8, 48, 50 Mean diffusivity changes seem to be less extensive, and display a progressive reversal as a function of the anatomical distance from the epileptogenic focus.46 Collectively, these abundant diffusion abnormalities imply a striking reconfiguration of white matter architecture in TLE, with changes being more prominent ipsilateral to the seizure focus, but also involving the contralateral hemisphere. More pronounced and diffuse disruptions in connectivity measures have been reported for left than right TLE.44, 48, 51 In addition, recent analyses indicated more extensive reduction of fiber density and cross-sectional area in TLE patients with hippocampal sclerosis compared with those not exhibiting global decreases in hippocampal volume.47





Figure 8.2. Network abnormalities in TLE. Exemplary findings of network alterations in TLE compared with controls are here displayed. Top left panel shows areas with aberrant cortical thickness correlations with medial thalamic volume. Top right panel displays fiber tracts with reduced fractional anisotropy. Bottom left panel shows abnormal fMRI time series (yellow) within a spatial component closely overlapping with the default mode network. Bottom right panel details results of a graph-theoretical analysis on interregional covariance of surface-based metrics (LH/RH = left/right hippocampus; LE/RE = left/right entorhinal cortex; LA/RA = left/right amygdala). Covariance increases/decreases are depicted in green/yellow.


(adapted from Bernhardt et al., Neurology 2012 with permission)

(adapted from Focke et al., NeuroImage 2008 with permission)

(adapted from Voets et al., Brain 2012 with permission)

(adapted from Bernhardt et al., Cereb Cortex 2015 with permission)

Extensive anomalies have also been inferred from functional connectivity measures. Resting-state fMRI studies found impaired connectivity estimates for mesiotemporal structures ipsilateral to the seizure focus, mostly involving links between anterior and posterior hippocampus, and between anterior hippocampus and entorhinal cortex.52, 53 Reduced functional connectivity was additionally detected between ipsilateral and contralateral hippocampus, ipsilateral and contralateral insula, and between epileptogenic mesiotemporal structures and bilateral lateral temporal neocortices.5456 Impaired ipsilateral functional integration may coexist with enhanced connectivity in contralateral mesiotemporal networks.52 It is suggested that more marked connectional derangements may take place in left than right TLE, both ipsilateral and contralateral to the seizure focus.54 At a whole-brain level, bilaterally impaired functional connectivity has been consistently detected for areas pertaining to the default mode network (DMN), which is traditionally composed of mesiotemporal lobes and mesial prefrontal, lateral, and midline parietal areas.5761 DMN activity is enhanced during task-free periods, and this network is thought to play a major role in internally directed activities, such as memory, future planning, and mind wandering.62 Interictal spike-correlated dysfunctions have also been mapped in regions belonging to the DMN by studies employing EEG-fMRI.63, 64 It remains poorly understood whether DMN integrity may be differentially affected according to the side of the epileptogenic focus. Hippocampal decoupling from anterior and posterior DMN nodes, however, appeared more prominent in patients with histology-confirmed hippocampal sclerosis, while being relatively less evident in those with gliosis only.65 Connectional derangements in TLE have also been documented for other large-scale functional systems, including sensory-motor, attentional, episodic memory, working memory, and language networks; altered functional integration has been furthermore described between mesiotemporal and subcortical structures, such as thalamus and cerebellum.5860, 6671 Disruption of thalamocortical functional connectivity were shown as ipsilateral in patients with focal seizures only, while involving both ipsilateral and contralateral structures in patients with additional secondarily generalized seizures.72 EEG-fMRI studies have indicated that activations correlated with temporal lobe spikes encompass a widespread ipsilateral network, most frequently extending to anterior hippocampus, amygdala, insula, superior temporal gyrus, basal ganglia, cerebellum, midcingulate as well as piriform cortex.73, 74 In view of its unique connectivity profile, the piriform cortex has been recently pinpointed as a common hub within networks involved in seizure generation in focal epilepsies, including TLE.75


Although neuroimaging-derived structural and functional abnormalities in TLE show considerable overlap, relatively few multimodal imaging studies have directly addressed cross-domain relationships. It is documented that decreased network integration of the hippocampus could be partially explained by estimates of its gray matter density.60, 76 More detailed structural analyses, however, suggest that the magnitude of hippocampal T2 signal changes, followed by hippocampal atrophy, may relate to the extent of its functional disconnection from anterior and posterior DMN epicenters.65 Moreover, disrupted functional connectivity between mesiotemporal structures and neocortical targets, including regions belonging to the DMN and sensory-motor networks, was also associated with altered diffusion parameters of the interconnecting white matter tracts.60, 61 In a recent study, abnormal functional amplitude metrics of midline and lateral default mode areas were shown to be mediated by microstructural abnormalities of the temporolimbic superficial white matter.77 Globally, these findings suggest that disruptions in morphological and architectural features may account for derangements in intrinsic functional connectivity in TLE.


Graph-theoretical studies on diffusion MRI datasets are suggestive of profound rearrangements within ipsilateral and contralateral mesiotemporal lobe subnetworks,8, 51, 78 which seem to account for a shift toward a more regularized topology.78 Analyses of functional data also found evidence of deranged limbic nodal topology55, 79 and detected changes indicative of compensatory reorganization of the contralateral mesiotemporal subnetwork.55 While aberrant topological properties were shown to maximally affect hippocampus, thalamus, and antero-mesial prefrontal cortex in TLE with hippocampal sclerosis, patients with normal hippocampal volumetry may only exhibit altered connectional properties of the ipsilateral temporal neocortex.80 Pronounced abnormalities have also been detailed by graph-theoretical studies addressing whole-brain network properties.81 An analysis on resting-state fMRI networks in bilateral TLE revealed decreased clustering and path length, along with a major redistribution of network hubs, which would indicate a random network arrangement.79 More recent resting-state fMRI work on unilateral TLE found increased clustering and path length, pointing to a more regular topology.82 A regularization of whole-brain network topology as well as pronounced shifts in the distribution of hubs and modularity was also reported by graph-theory studies on structural MRI datasets, such as cortical thickness or gray matter volume correlations83, 84 and diffusion MRI metrics,85, 86 and by graph-theoretical analyses on EEG-derived networks.87, 88 Evidence pointing to a regularization of whole-brain networks also came from a recent meta-analysis, which included electrophysiology and imaging graph-theory studies on focal epilepsy cohorts.89 Graph-theoretical studies also indicated reduced coupling between structural and functional networks, which may be partially modulated by disease duration.90 Connectome alterations appeared to be more pervasive in left TLE than right TLE patients.51, 84



8.5 Emerging Evidence for Disturbed Connectivity in Extratemporal Epilepsies


Extratemporal epilepsy related to cortical malformations, particularly FCD, is a major cause of drug-resistant seizures. In recent years, advances in MRI processing have led to substantial improvements in lesion detection.91 In parallel, several studies have unveiled the existence of morphological and architectural abnormalities, encompassing gray matter density, cortical thickness, sulcal depth, diffusion parameters, and local functional markers in areas at a distance from the dysplastic cortex.9297 These findings have prompted further attempts to assess the integrity of structural and functional networks in epilepsy secondary to FCD.


Diffusion MRI studies in patients with FCD-related frontal lobe epilepsy (FLE) found bilateral reductions in fractional anisotropy, encompassing the superior longitudinal fasciculus, uncinate fasciculus, cingulum, as well as corpus callosum.92, 98100 Bilateral diffusion abnormalities involving the corpus callosum and several interlobar fiber tracts are also described for pediatric FLE patients with nondiagnostic MRI.101, 102 Widespread diffusivity increases are also reported for bilateral frontal, temporal, and parietal lobes and corpus callosum.99, 103 Only a few fMRI studies probed the integrity of functional networks in homogeneous populations of patients with FCD. Functional connectivity analyses seeded from dysplastic cortices revealed complex rearrangements of network properties, involving variable combinations of hyper- and hypoconnectivity.104 There is also evidence of higher signal variability of local resting-state fMRI measures in a sample of patients with extratemporal epilepsy, half of whom with FCDs.105 Striking reconfigurations of language and memory networks, with evidence of intra- and interhemispheric redistribution of function, have been detailed for FLE and mixed focal epilepsy patients.106109 In adult and pediatric populations with FLE and suspected dysplasia, resting-state fMRI analyses have displayed extensive connectional derangements, involving language,110 working memory,111, 112 sensory-motor networks,113 as well as DMN, attentional, visual, and auditory networks.114 Focusing on patients with malformations of cortical development or pure dysplasia cohorts, EEG-fMRI studies have shown widely distributed spike-correlated BOLD signal changes, frequently involving cortical and subcortical areas at a distance from the seizure onset zone.73, 115117 These studies are indicative of widespread epileptogenic networks, which may also imply increased functional connectivity between the epileptogenic region and remote brain areas, possibly with patient-specific connectional profiles.118 A graph-theoretical study conducted on resting-state fMRI data in adult patients with nonlesional focal epilepsy found evidence for a reduced global network efficiency as well as decreased clustering coefficient, suggestive of a network randomization.119 Conversely, metrics indicative of a more regularized network architecture were reported by resting-state fMRI analyses on a mixed adult sample of FCD-related and MRI-negative extratemporal epilepsy,120 and in a small group of adults with polymicrogyria.121 Results of DTI and resting-state fMRI analyses in pediatric populations with nonlesional FLE are also compatible with a more regularized network topology.122, 123



8.6 Network Disturbances in Idiopathic Generalized Epilepsies


Challenging conventional beliefs, several studies carried out in the last decade have provided compelling evidence that structural abnormalities are indeed present in patients with IGE, and most frequently encompass thalamus and frontal cortical areas.2, 124 In parallel, the pivotal role of activity changes in the thalamus and diffuse cortical areas during generalized spike-wave discharges was documented by EEG-fMRI analyses.125 As a result, a wealth of investigations has recently assessed the integrity of structural and functional networks in IGE, with prominent attention to the interplay between thalamus and neocortical areas (Figure 8.3).





Figure 8.3. Network abnormalities in IGE. Exemplary findings of network alterations in IGE compared with controls are here depicted. Top panel highlights regions with increased cortical thickness correlations with thalamic volume in patients with IGE-GTCS. In the middle panel, increases (warm colors) and decreases (cold colors) in DTI connectivity of the pre-SMA/SMA are displayed in the upper/lower row. Lower panel displays EEG-correlated fMRI activations (warm colors) and deactivations (cold colors) in a mixed IGE cohort.


(adapted from Bernhardt et al., NeuroImage 2009 with permission)

(adapted from Vollmar et al., Neurology 2012 with permission)

(adapted from Gotman et al., with permission, copyright (2005) National Academy of Sciences, USA)

In a patient cohort of IGE with generalized tonic-clonic seizures (IGE-GTCS), one of the first studies to assess the integrity of thalamocortical networks using MRI covariance reported enhanced thalamic structural coupling with frontocentral, parietal, and lateral temporal cortices, along with connectivity reductions between thalamus and limbic areas.126 Studies employing diffusion MRI have largely focused on juvenile myoclonic epilepsy (JME), and consistently indicate impaired thalamo-frontocortical connectivity, particularly in the frontal lobe.127130 White matter abnormalities in JME, however, also encompass associative fiber tracts, such as the superior longitudinal fasciculus, the uncinate fasciculus,131, 132 and segments of the corpus callosum connecting prefrontal cortices, supplementary motor areas (SMA), and posterior cingulate cortices.130, 133135 Vollmar and colleagues recently described a complex pattern of diffusion derangements in JME, including increased structural connectivity of pre-SMA with motor cortices and descending motor pathways, and reduced connectivity between pre-SMA and prefrontal/frontopolar areas, which may represent a correlate of the impaired frontal lobe functions observed in this syndrome. In addition, enhanced connectivity between SMA and the occipital lobe as well as lateral temporal neocortices was also described.136 A graph theoretical analysis on a diffusion MRI dataset indicated increased structural connectivity in a subnetwork including primary motor cortex, parietal lobe areas, right hippocampus, and subcortical structures such as putamen and cerebellum.137 Collectively, these findings are suggestive of complex reconfigurations affecting white matter networks in JME. Diffusion imaging studies on mixed IGE cohorts also found evidence of disrupted white matter microstructure, involving anterior thalamocortical fibers, corticospinal tracts, and several cortico-cortical projections;138, 139 a subgroup analysis failed to detect major differences between JME patients and the remainder mixed IGE group.138 Preliminary evidence of diffusion abnormalities extending to cortico-cortical and corticosubcortical fibers is also available for childhood absence epilepsy (CAE).140, 141


Studies employing EEG-fMRI have led to major improvements in our understanding of functional networks implicated in generalized spike-wave discharges and absence seizures. Discharge-correlated BOLD activations are consistently observed in thalamus and frontal lobes, along with deactivations in a collection of areas pertaining to the DMN.125, 142146 Other analyses additionally emphasized the contribution of deactivations in basal ganglia, such as the caudate.147149 Thalamocortical BOLD activations are hypothesized as underlying the generation of the epileptic paroxysms, in line with evidence from experimental and animal models. On the other hand, perturbation of the brain’s default state is thought to arise as an indirect consequence of these discharges,125 although more recent evidence points to a permissive role of precuneal activity, which would act as the initiator of spike-wave discharges.150 Studies focusing on the dynamic changes taking place during absence seizures revealed that cortical activity within a set of frontal and parietal regions precedes thalamic activation,151, 152 and that frontal activation may exhibit a causal relationship to subsequent thalamic changes.153 While BOLD dynamic changes exhibit high variability across subjects, consistent within-subject results across multiple epileptic events were observed.152 These findings are in line with results of electrical source mapping studies, which emphasize focal patterns of ictal onsets, mostly involving mesial frontal and orbitofrontal cortices,11, 154 and would revive the concept of a “cortical focus” as the initiator of generalized spike-wave discharges.10 More recent EEG-fMRI work, however, identified activity changes within multiple distinct subnetworks during absence seizures, and suggests that early discharge-associated changes may involve an extended network encompassing medial parietal, lateral occipital, frontopolar areas as well as subcortical structures.155


Resting-state fMRI assessments have reported a widespread set of connectional derangements involving the DMN in CAE, IGE-GTCS, and mixed IGE cohorts.156159 Pairwise functional connectivity between DMN nodes is usually reported as being decreased. On the other hand, enhanced connectivity has been observed between mesial prefrontal cortex and regions implicated in “task-positive” cortical regions, including the superior parietal lobule and the intraparietal sulcus, and between medial parieto-occipital cortices and primary sensory networks, suggesting a reduced segregation of the DMN.157, 160 Increased interhemispheric connectivity is described for orbitofrontal cortices in CAE,161 and for anterior cingulate and mesial prefrontal cortices in IGE-GTCS.162 Additional abnormalities in CAE include decreased functional connectivity between anterior insular/frontal opercular and mesial prefrontal cortices,163 which may indicate impaired functioning of the attentional network in this syndrome, as well as abnormal connectional profiles between thalamus and diffuse cortical areas.160 The numerous assessments of thalamocortical functional connectivity patterns, however, do not exhibit homogeneous results. In patients with IGE-GTCS, decreased resting-state functional connectivity was described between the medial dorsal thalamic nucleus and bilateral orbitofrontal cortices, amygdalae, and basal ganglia.164 Studies on mixed IGE groups identified increased functional connectivity between posterior thalamus and mesial prefrontal/anterior cingulate cortices.165 Opposite results for the same cortical targets were instead obtained when seeding from anteromedial thalamic subsections,166 which seem to exhibit reduced amplitude of low-frequency fluctuations.167 Connectivity between thalamus and posteromedial default mode areas, such as the posterior cingulate cortex, was uniformly reported as decreased.165, 166 Finally, a recent study assessing a heterogeneous IGE population documented increased functional connectivity measures for a majority of the analyzed thalamocortical connections, including thalamotemporal, prefrontal-thalamic, occipitothalamic, as well as motor and premotor thalamic networks.168


Studies assessing patients with JME and their unaffected siblings revealed enhanced functional connectivity between motor and prefrontal cortices, which accounts for the coactivation of the motor system during cognitive tasks described in this syndrome and represents a disease endophenotype.169, 170 Such functional connectional patterns are also mirrored by increased structural connectivity of the underlying white matter tracts in JME, as reported by diffusion MRI studies.136 Additional alterations in JME include decreased functional connectivity between thalamus and the SMA, which parallels abnormal diffusivity parameters of the interconnecting white matter projections.129


Graph theoretical findings in IGE cohorts are, so far, relatively heterogeneous. A diffusion MRI study on children with CAE reported whole-brain topological rearrangement, indicated by reduced clustering coefficient and increased path length,141 while a resting-state fMRI analysis indicated aberrant centrality measures of network hubs, including fronto-temporoparietal DMN nodes as well as thalamus.171 Conversely, no major differences between adult patients with JME and controls were found for whole-brain diffusion MRI-derived graph theoretical metrics.137 In IGE-GTCS, combined graph-theory approaches on resting-state fMRI and diffusion MRI datasets indicated disruptions in whole-brain network topology, with a shift toward a random network configuration in both domains.172 Across both imaging modalities, major derangements in nodal characteristics172 were particularly evident for key corticosubcortical network hubs, pointing to an impaired “rich club” organization.173 There was also evidence for decoupling between structural and functional connectivity networks, which was shown to be negatively associated with epilepsy duration and may thus underlie disease progression.172



8.7 Relevance of Network Analysis for Clinical Purposes


In epilepsy, the majority of studies have employed MRI-derived connectivity measures to describe functional and structural correlates of disease processes. Increasing evidence indicates that network analyses may also represent promising tools to inform clinical practice, especially with respect to the management of focal epilepsy. Several studies have assessed the potential of resting-state fMRI and diffusion MRI analyses to lateralize the seizure focus. Patterns of thalamotemporal functional connectivity were shown to categorize left and right TLE patients with high sensitivity and specificity,53, 174 whereas connectivity increases in contralateral limbic structures may aid the identification of the nonepileptic temporal lobe.175 A lateralization index derived from the analysis of local functional markers, such as the amplitude of the low-frequency fluctuations (ALFF), seems to effectively distinguish left from right TLE.176 Successful detection of the seizure focus could also be attained by contrasting the local functional connectivity profiles within lesional areas to that of contralateral corresponding regions.177179 Implementing functional topology metrics within an automated algorithm resulted in prospective prediction of focus laterality with similar accuracy rates compared to video-EEG, and with a significantly improved yield compared to expert qualitative MRI inspection.180


Initial evidence indicates that structural and functional network modeling may provide valuable indications regarding disease monitoring and prognosis. A longitudinal analysis based on cortical thickness correlations showed an increase in path length over time, suggesting that serial analyses of network topology metrics might capture changes potentially related to disease progression.83 Graph-theoretical analyses in a sample of TLE patients who underwent epilepsy surgery revealed increased network resilience metrics in individuals rendered seizure-free compared to those with poor outcome.181 Moreover, a longitudinal comparison of presurgical with postsurgical connectome-level metrics documented a reorganization of connections involving mesial prefrontal cortices and temporoparietal junctions, which appeared more pronounced in seizure-free patients.181 In a diffusion MRI study in TLE, disrupted organizational properties within the ipsilateral temporal subnetwork and increased structural connectivity between a set of ipsilateral and contralateral temporoparietal areas were associated with poor postsurgical seizure control.8 Recent analyses in TLE have also emphasized the relevance of thalamic connectivity. Bilateral thalamic atrophy and abnormal DTI-derived thalamotemporal probabilistic paths were distinctive features of patients with persistent seizures after surgery compared with those rendered seizure-free.50 In a resting-state fMRI analysis, patients with poor postsurgical seizure outcome presented with higher values of centrality (“hubness”) metrics within ipsilateral and contralateral thalami, along with enhanced connectivity of both thalami with bilateral cortical targets.182 Computing whole-brain functional connectivity from seizure foci identified with EEG-fMRI, Negishi and colleagues found that high functional connectivity within the epileptogenic hemisphere may be a predictor of successful seizure control after surgery.183 The potential of interhemispheric functional connectivity differences to distinguish patients with good and suboptimal surgical outcome is also reported by other resting-state fMRI work.184 Recent studies have attempted to identify connectome-based biomarkers to predict outcome at the individual level. Embedding preoperative white matter architectural features into classification algorithms led to a fairly accurate prediction of postsurgical outcomes in TLE,185, 186 which displayed higher sensitivity and specificity than a combined scored derived from clinical characteristics.186 Machine-learning models based on thalamic connectional properties, as derived by resting-state fMRI graph-theoretical analysis, were recently shown to provide a more accurate prediction of postsurgical outcome than algorithms based on clinical variables.182


Network analysis may prove useful for deriving functional or structural biomarkers of cognitive functions. Resting-state functional connectivity maps obtained from seeds placed in the inferior frontal cortices predicted hemispheric dominance for language in TLE.187 Functional connectivity between areas belonging to frontotemporal language networks was demonstrated to be associated with verbal IQ measures188 and with scores in word fluency and text reading tasks.110 Several studies have also found imaging correlates of verbal and nonverbal episodic memory. In TLE, functional connectivity disruptions between mesiotemporal lobes and posterior DMN areas may underlie impairments in episodic memory performance.76, 189, 190 Enhanced functional connectivity between the epileptogenic mesiotemporal lobe and DMN regions, however, is elsewhere reported to correlate with worse memory abilities.191, 192 Recruitment of contralateral mesiotemporal and DMN structures might instead relate to improved memory performance.52, 191, 192 Measures of bilateral functional connectivity between the thalamus and neocortical areas, including prefrontal cortex and supramarginal gyrus, may be associated with working memory scores.71 Of note, diffusion abnormalities in specific fiber tracts, such as uncinate, arcuate, inferior longitudinal fascicles and cingulum, were shown to correlate with memory and language abilities,193195 and may be used to predict scores in verbal memory tests.196 In populations of adult and pediatric localization-related epilepsy patients, graph theoretical metrics derived from fMRI119, 122 and diffusion MRI datasets197 displayed the ability to mirror cognitive performance across several domains.


Connectivity metrics may assist the prediction of neurocognitive outcome after epilepsy surgery. Postsurgical episodic memory performance was shown to correlate negatively with functional connectivity between the epileptogenic hippocampus and precunei, and positively with connectional profiles between the contralateral hippocampus and precunei.189 A combination of diffusion MRI, resting-state as well as task-based fMRI was recently shown to accurately predict verbal fluency outcomes following anterior temporal lobe surgery.198 Moreover, an analysis on resting-state fMRI data points to the usefulness of regional graph theory measures in informing predictive models of postsurgical cognitive outcome across multiple domains, including language, attention, working memory, and verbal episodic memory.199 Interestingly, enhanced whole-brain integration of the contralateral unaffected hippocampus also predicted a better cognitive outcome,199 adding further proof that recruitment of contralateral areas may represent a compensatory phenomenon favoring memory abilities.



8.8 Conclusions


Nowadays, a network perspective is essential to the understanding of the development, progression, and management of epilepsy. Due to recent advancements in noninvasive imaging, networks can be mapped at multiple levels, spanning from local and interregional connectivity to whole-brain topological attributes, thus providing a window into the complex pattern of disease-specific effects. Functional and structural MRI analyses have contributed to substantial improvements in our understanding of brain connectivity and have challenged the originally proposed dichotomy between focal and generalized syndromes. In focal epilepsies, there is evidence for organizational derangements at a whole-brain level. In IGE, although reconfigurations of white matter architecture and shifts in network topology are described, several studies do not support a homogeneous whole-brain involvement, and rather emphasize the role of distinct networks. It is anticipated that the rich set of connectivity metrics and topological properties can serve both diagnostics and prognostics in individual patients. Caution is required before their clinical implementation, which will need to be preceded by validation on large, multicenter cohorts with assessments of reliability, sensitivity, and specificity, and formulation of guidelines pertaining to data acquisition and analysis.




References


1.Spencer SS. Neural networks in human epilepsy: evidence of and implications for treatment. Epilepsia. 2002;43:219–27. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

2.Richardson MP. Large scale brain models of epilepsy: dynamics meets connectomics. J Neurol Neurosurg Psychiatry. 2012;83:1238–48. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

3.Bernhardt BC, Hong S, Bernasconi A, Bernasconi N. Imaging structural and functional brain networks in temporal lobe epilepsy. Front Hum Neurosci. 2013;7:624. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

4.Caciagli L, Bernhardt BC, Hong SJ, Bernasconi A, Bernasconi N. Functional network alterations and their structural substrate in drug-resistant epilepsy. Front Neurosci. 2014;8:411. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

5.Tellez-Zenteno JF, Dhar R, Wiebe S. Long-term seizure outcomes following epilepsy surgery: a systematic review and meta-analysis. Brain. 2005;128:1188–98. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

6.de Tisi J, Bell GS, Peacock JL, et al. The long-term outcome of adult epilepsy surgery, patterns of seizure remission, and relapse: a cohort study. Lancet. 2011;378:1388–95. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

7.Bernhardt BC, Bernasconi N, Concha L, Bernasconi A. Cortical thickness analysis in temporal lobe epilepsy: reproducibility and relation to outcome. Neurology. 2010;74:1776–84. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar

8.Bonilha L, Helpern JA, Sainju R, et al. Presurgical connectome and postsurgical seizure control in temporal lobe epilepsy. Neurology. 2013;81:1704–10. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

9.Blumenfeld H. Cellular and network mechanisms of spike-wave seizures. Epilepsia. 2005;46(suppl 9):2133. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

10.Meeren HK, Pijn JP, Van Luijtelaar EL, Coenen AM, Lopes da Silva FH. Cortical focus drives widespread corticothalamic networks during spontaneous absence seizures in rats. J Neurosci. 2002;22:1480–95. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

11.Holmes MD, Brown M, Tucker DM. Are “generalized” seizures truly generalized? Evidence of localized mesial frontal and frontopolar discharges in absence. Epilepsia. 2004;45:1568–79. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

12.Carney PW, Jackson GD. Insights into the mechanisms of absence seizure generation provided by EEG with functional MRI. Front Neurol. 2014;5:162. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

13.Amor F, Baillet S, Navarro V, Adam C, Martinerie J, Quyen Mle V. Cortical local and long-range synchronization interplay in human absence seizure initiation. NeuroImage. 2009;45:950–62. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

14.Wolf P, Yacubian EM, Avanzini G, et al. Juvenile myoclonic epilepsy: a system disorder of the brain. Epilepsy Res. 2015;114:212. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar

15.Avanzini G, Manganotti P, Meletti S, et al. The system epilepsies: a pathophysiological hypothesis. Epilepsia. 2012;53:771–8. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

16.Power JD, Fair DA, Schlaggar BL, Petersen SE. The development of human functional brain networks. Neuron. 2010;67:735–48. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

17.Sporns O, Tononi G, Kotter R. The human connectome: a structural description of the human brain. PLOS Comput Biol. 2005;1:e42. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

18.Mori S, Crain BJ, Chacko VP, van Zijl PC. Three-dimensional tracking of axonal projections in the brain by magnetic resonance imaging. Ann Neurol. 1999;45:265–9. CrossRef | <A title="Find at Chinese University of Hong Kong" class=resolver-link href="julac.hosted.exlibrisgroup.com/openurl/CUHK_ALMA/CUHK_SERVICES_PAGE?genre=article&date=1999&doi=10.1002/1531-8249(199902)45:23.0.CO;2-3&sid=www.cambridge.org&title=Ann%20Neurol&atitle=Three-dimensional%20tracking%20of%20axonal%20projections%20in%20the%20brain%20by%20magnetic%20resonance%20imaging&jtitle=Ann%20Neurol&volume=45” target=_blank alt=”Chinese University of Hong Kong”>Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

19.Behrens TE, Johansen-Berg H, Woolrich MW, et al. Non-invasive mapping of connections between human thalamus and cortex using diffusion imaging. Nat Neurosci. 2003;6:750–7. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

20.Le Bihan D, Mangin JF, Poupon C, et al. Diffusion tensor imaging: concepts and applications. J Magn Reson Imaging. 2001;13:534–46. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

21.Conturo TE, Lori NF, Cull TS, et al. Tracking neuronal fiber pathways in the living human brain. Proc Natl Acad Sci USA. 1999;96:10422–7. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

22.Knosche TR, Anwander A, Liptrot M, Dyrby TB. Validation of tractography: Comparison with manganese tracing. Hum Brain Mapp. 2015;36:4116–34. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

23.Jones DK, Knosche TR, Turner R. White matter integrity, fiber count, and other fallacies: the do’s and don’ts of diffusion MRI. NeuroImage. 2013;73:239–54. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

24.Alexander-Bloch A, Giedd JN, Bullmore E. Imaging structural co-variance between human brain regions. Nat Rev Neurosci. 2013;14:322–36. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

25.Alexander-Bloch A, Raznahan A, Bullmore E, Giedd J. The convergence of maturational change and structural covariance in human cortical networks. J Neurosci. 2013;33:2889–99. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

26.Gong G, He Y, Chen ZJ, Evans AC. Convergence and divergence of thickness correlations with diffusion connections across the human cerebral cortex. NeuroImage. 2012;59:1239–48. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

27.Kelly C, Biswal BB, Craddock RC, Castellanos FX, Milham MP. Characterizing variation in the functional connectome: promise and pitfalls. Trends Cogn Sci. 2012;16:181–8. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

28.Clos M, Rottschy C, Laird AR, Fox PT, Eickhoff SB. Comparison of structural covariance with functional connectivity approaches exemplified by an investigation of the left anterior insula. NeuroImage. 2014;99:269–80. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

29.Lerch JP, Worsley K, Shaw WP, et al. Mapping anatomical correlations across cerebral cortex (MACACC) using cortical thickness from MRI. NeuroImage. 2006;31:9931003. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar

30.Zielinski BA, Gennatas ED, Zhou J, Seeley WW. Network-level structural covariance in the developing brain. Proc Natl Acad Sci USA. 2010;107:18191–6. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

31.Friston KJ. Functional and effective connectivity: a review. Brain Connect. 2011;1:1336. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

32.Cabral J, Kringelbach ML, Deco G. Exploring the network dynamics underlying brain activity during rest. Prog Neurobiol. 2014;114:102–31. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar

33.Damoiseaux JS, Rombouts SA, Barkhof F, et al. Consistent resting-state networks across healthy subjects. Proc Natl Acad Sci USA. 2006;103:13848–53. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

34.Centeno M, Carmichael DW. Network connectivity in epilepsy: resting state fMRI and EEG-fMRI contributions. Front Neurol. 2014;5:93. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

35.Bullmore E, Sporns O. Complex brain networks: graph theoretical analysis of structural and functional systems. Nat Rev Neurosci. 2009;10:186–98.Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

36.Watts DJ, Strogatz SH. Collective dynamics of “small-world” networks. Nature. 1998;393:440–2. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

37.Bullmore E, Sporns O. The economy of brain network organization. Nat Rev Neurosci. 2012;13:336–49. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

38.Bernhardt BC, Worsley KJ, Besson P, et al. Mapping limbic network organization in temporal lobe epilepsy using morphometric correlations: insights on the relation between mesiotemporal connectivity and cortical atrophy. NeuroImage. 2008;42:515–24. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

39.Bonilha L, Rorden C, Halford JJ, et al. Asymmetrical extra-hippocampal grey matter loss related to hippocampal atrophy in patients with medial temporal lobe epilepsy. J Neurol Neurosurg Psychiatry. 2007;78:286–94.Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

40.Mueller SG, Laxer KD, Barakos J, Ian C, Garcia P, Weiner MW. Widespread neocortical abnormalities in temporal lobe epilepsy with and without mesial sclerosis. NeuroImage. 2009;46:353–9. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

41.Mueller SG, Laxer KD, Barakos J, et al. Involvement of the thalamocortical network in TLE with and without mesiotemporal sclerosis. Epilepsia. 2010;51:1436–45.Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

42.Bernhardt BC, Bernasconi N, Kim H, Bernasconi A. Mapping thalamocortical network pathology in temporal lobe epilepsy. Neurology. 2012;78:129–36. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

43.Concha L, Beaulieu C, Collins DL, Gross DW. White-matter diffusion abnormalities in temporal-lobe epilepsy with and without mesial temporal sclerosis. J Neurol Neurosurg Psychiatry. 2009;80:312–9. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

44.Focke NK, Yogarajah M, Bonelli SB, Bartlett PA, Symms MR, Duncan JS. Voxel-based diffusion tensor imaging in patients with mesial temporal lobe epilepsy and hippocampal sclerosis. NeuroImage. 2008;40:728–37. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

45.Thivard L, Lehericy S, Krainik A, et al. Diffusion tensor imaging in medial temporal lobe epilepsy with hippocampal sclerosis. NeuroImage. 2005;28:682–90. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

46.Concha L, Kim H, Bernasconi A, Bernhardt BC, Bernasconi N. Spatial patterns of water diffusion along white matter tracts in temporal lobe epilepsy. Neurology. 2012;79:455–62. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

Only gold members can continue reading. Log In or Register to continue

Jan 3, 2021 | Posted by in NEUROLOGY | Comments Off on Chapter 8 – Network Modeling of Epilepsy Using Structural and Functional MRI
Premium Wordpress Themes by UFO Themes