Figure 3 - available via license: Creative Commons Attribution 3.0 Unported
Content may be subject to copyright.
Contrast and sharpness measures encode related but different matrix features. Analysis of example matrices with different levels of Gaussian smoothing (left vs right panels), and different signal amplitudes (top vs bottom panels)
Source publication
Electroencephalography (EEG) allows recording of cortical activity at high temporal resolution. EEG recordings can be summarised along different dimensions using network-level quantitative measures, e.g. channel-to-channel correlation, or band power distributions across channels. These reveal network patterns that unfold over a range of different t...
Similar publications
Electroencephalography (EEG) allows recording of cortical activity at high temporal resolution. EEG recordings can be summarised along different dimensions using network-level quantitative measures, e.g. channel-to-channel correlation, or band power distributions across channels. These reveal network patterns that unfold over a range of different t...
Citations
... Notably, these stable modes of activity are activated in a sequential manner, and analysing the sequences of activation can provide valuable insights into the overall dynamics of the brain on a large scale. Analysing brain network dynamics -especially in resting state -has proved useful, for instance in pain research, in epilepsy research Rosch et al., 2018a) and in Alzheimer's research (Kuang et al., 2019;Núñez et al., 2021;Smailovic et al., 2019). ...
A pervasive challenge in neuroscience is testing whether neuronal connectivity changes over time due to specific causes, such as stimuli, events, or clinical interventions. Recent hardware innovations and falling data storage costs enable longer, more naturalistic neuronal recordings. The implicit opportunity for understanding the self-organised brain calls for new analysis methods that link temporal scales: from the order of milliseconds over which neuronal dynamics evolve, to the order of minutes, days or even years over which experimental observations unfold. This review article demonstrates how hierarchical generative models and Bayesian inference help to characterise neuronal activity across different time scales. Crucially, these methods go beyond describing statistical associations among observations and enable inference about underlying mechanisms. We offer an overview of fundamental concepts in state-space modeling and suggest a taxonomy for these methods. Additionally, we introduce key mathematical principles that underscore a separation of temporal scales, such as the slaving principle, and review Bayesian methods that are being used to test hypotheses about the brain with multi-scale data. We hope that this review will serve as a useful primer for experimental and computational neuroscientists on the state of the art and current directions of travel in the complex systems modelling literature.
... The incidence rate of epilepsy is particularly high in infancy and childhood. The characteristics of early infant EEG are various spatially distributed activities, rather than the more typical posterior rhythm in the mature EEG (Rosch et al., 2018). In addition, the electrographic symptoms of seizures in children are not as typical as those in adults. ...
Changes in the frequency composition of the human electroencephalogram are associated with the transitions to epileptic seizures. Cross-frequency coupling (CFC) is a measure of neural oscillations in different frequency bands and brain areas, and specifically phase–amplitude coupling (PAC), a form of CFC, can be used to characterize these dynamic transitions. In this study, we propose a method for seizure detection and prediction based on frequency domain analysis and PAC combined with machine learning. We analyzed two databases, the Siena Scalp EEG database and the CHB-MIT database, and used the frequency features and modulation index (MI) for time-dependent quantification. The extracted features were fed to a random forest classifier for classification and prediction. The seizure prediction horizon (SPH) was also analyzed based on the highest-performing band to maximize the time for intervention and treatment while ensuring the accuracy of the prediction. Under comprehensive consideration, the results demonstrate that better performance could be achieved at an interval length of 5 min with an average accuracy of 85.71% and 95.87% for the Siena Scalp EEG database and the CHB-MIT database, respectively. As for the adult database, the combination of PAC analysis and classification can be of significant help for seizure detection and prediction. It suggests that the rarely used SPH also has a major impact on seizure detection and prediction and further explorations for the application of PAC are needed.
... A higher C likely reflects increased short-range connectivity, and increased BC reflects robust broad network. Higher frequency bands are more involved in establishing cognitive representation, whereas lower frequencies are more anatomically constrained (Rosch et al., 2018). The network in the developing infantile brain changes as it is dynamically modulated by local microcircuit and global network integration (Basset and Bullmore, 2006). ...
Objective
This study aimed to determine the correlation between outcomes following adrenocorticotrophic hormone (ACTH) therapy and measurements of relative power spectrum (rPS), weighted phase lag index (wPLI), and graph theoretical analysis on pretreatment electroencephalography (EEG) in infants with non-lesional infantile epileptic spasms syndrome (IESS).
Methods
Twenty-eight patients with non-lesional IESS were enrolled. Outcomes were classified based on seizure recurrence following ACTH therapy: seizure-free (F, n=21) and seizure-recurrence (R, n=7) groups. The rPS, wPLI, clustering coefficient, and betweenness centrality were calculated on pretreatment EEG and were statistically analyzed to determine the correlation with outcomes following ACTH therapy.
Results
The rPS value was significantly higher in the delta frequency band in group R than in group F (p<0.001). The wPLI values were significantly higher in the delta, theta, and alpha frequency bands in group R than in group F (p=0.007, <0.001, and <0.001, respectively). The clustering coefficient in the delta frequency band was significantly lower in group R than in group F (p<0.001).
Conclusions
Our findings demonstrate the significant differences in power and functional connectivity between outcome groups.
Significance
This study may contribute to an early prediction of ACTH therapy outcomes and thus help in the development of appropriate treatment strategies.
... It is widely believed that GRE is caused by the invasion and destruction of subcortical structures, leading to a whole brain functional disorder [9,11]. Previous research has confirmed multiple global network attribute changes in patients with epilepsy [12][13][14]. The structural networks are also involved and are related to abnormalities in the main white matter fiber tracts in patients with epilepsy [15,16]. ...
(1) Background: Glioma is the most common primary tumor in the central nervous system, and glioma-related epilepsy (GRE) is one of its common symptoms. The abnormalities of white matter fiber tracts are involved in attributing changes in patients with epilepsy (Rudà, R, 2012).This study aimed to assess frontal lobe gliomas’ effects on the cerebral white matter fiber tracts. (2) Methods: Thirty patients with frontal lobe glioma were enrolled and divided into two groups (Ep and nEep). Among them, five patients were excluded due to apparent insular or temporal involvement. A set of 14 age and gender-matched healthy controls were also included. All the enrolled subjects underwent preoperative conventional magnetic resonance images (MRI) and diffusion tensor imaging (DTI). Furthermore, we used tract-based spatial statistics to analyze the characteristics of the white matter fiber tracts. (3) Results: The two patient groups showed similar patterns of mean diffusivity (MD) elevations in most regions; however, in the ipsilateral inferior fronto-occipital fasciculus (IFOF), superior longitudinal fasciculus (SLF), and superior corona radiata, the significant voxels of the EP group were more apparent than in the nEP group. No significant fractional anisotropy (FA) elevations or MD degenerations were found in the current study. (4) Conclusions: Gliomas grow and invade along white matter fiber tracts. This study assessed the effects of GRE on the white matter fiber bundle skeleton by TBSS, and we found that the changes in the white matter skeleton of the frontal lobe tumor-related epilepsy were mainly concentrated in the IFOF, SLF, and superior corona radiata. This reveals that GRE significantly affects the white matter fiber microstructure of the tumor.
... The classification of such transitions is presently incomplete, but we will discuss an example below. We assume that such classification will necessarily involve advanced multivariate data analysis and will benefit from machine learning techniques as in the case of abnormally intermittent activity related to newborn and early childhood epileptic encephalopathies (Rosch et al. 2017). ...
... One step further and transcending network analysis, strategies can also be built to classify patients with epilepsy and controls with raw imaging and neurophysiological datasets where no data is predefined or no labels are available (Roy et al., 2019). Thirdly, since the brain is in a constant state of change, characterization of the brain network organization does not only require the calculation of static network metrics but may benefit from taking the dynamic properties into account (Pedersen et al., 2017;Rosch et al., 2018). Combining the dynamics of brain regions and their interconnectedness could lead to a more comprehensive understanding of epilepsy-associated network abnormalities. ...
Abnormalities of the brain network organization in focal epilepsy have been extensively quantified. However, the extent and directionality of abnormalities are highly variable and subtype insensitive. We conducted meta-analyses to obtain a more accurate and epilepsy type-specific quantification of the interictal global brain network organization in focal epilepsy. By using random-effects models, we estimated differences in average clustering coefficient, average path length, and modularity between patients with focal epilepsy and controls, based on 45 studies with a total sample size of 1,468 patients and 1,021 controls. Structural networks had a significant lower level of integration in patients with epilepsy as compared to controls, with a standardized mean difference of -0.334 (95% confidence interval -0.631 to -0.038; p-value 0.027). Functional networks did not differ between patients and controls, except for the beta band clustering coefficient. Our meta-analyses show that differences in the brain network organization are not as well defined as individual studies often propose. We discuss potential pitfalls and suggestions to enhance the yield and clinical value of network studies.
... This paper introduces a class of dynamic causal model (DCM) that can be used for characterising slow fluctuations in biophysical parameters that might underlie phase transitions in the brain. This method is based on a separation of temporal scales ( Jirsa et al., 1994 ;Papadopoulou et al., 2017 ;Rosch et al., 2018a ;Rosch et al., 2018b ;Rosch et al., 2018c ;Blenkinsop et al., 2012 ;Jirsa et al., 2014 ;Steyn-Ross and Steyn-Ross, 2010 ;Nevado-Holgado et al., 2012 ) where fast neuronal fluctuations are generated by slow fluctuations in synaptic parameters and other neurophysiological parameters (e.g., extracellular potassium). DCM then allows one to specify different hypotheses about causal relations between slow biological mechanisms ( Papadopoulou et al., 2015 ) and select the most likely model that explains phase transitions in electrophysiological data. ...
... This means that one can specify an adiabatic DCM to model slow dynamics such as spike rate adaptation, short-term plasticity or, indeed, the target of this work; trajectories in parameter space that engender paroxysmal transitions in neuronal dynamics, e.g., epilepsy. The basic DCM for CSD using this work has been described in many previous applications e.g., ( Papadopoulou et al., 2017 ;Rosch et al., 2018a ;Rosch et al., 2018b ). The key thing that we bring to the table is equipping the model with a second level that is constrained by empirical data features at a slower timescale. ...
... This is because crossing a phase boundary or 'separatrix' in parameter space induces the transitions in mesoscopic activity. Although there is a sudden change in the spectral activity induced by this boundary crossing, the drift of the parameters per se is quite smooth and slow (in comparison with the fast neuronal states) ( Papadopoulou et al., 2017 ;Rosch et al., 2018a ;Rosch et al., 2018c ). The synchronised beta burst is a hallmark of movement disorders ( McCarthy et al., 2011 ;Spitzer and Haegens, 2017 ;Sherman et al., 2016 ), which may also be induced by drugs/interventions ( Rodriguez et al., 2004 ;Shin et al., 2017 ) or during memory retrieval ( Jansen et al., 2011 ). ...
This technical note introduces adiabatic dynamic causal modelling, a method for inferring slow changes in biophysical parameters that control fluctuations of fast neuronal states. The application domain we have in mind is inferring slow changes in variables (e.g., extracellular ion concentrations or synaptic efficacy) that underlie phase transitions in brain activity (e.g., paroxysmal seizure activity). The scheme is efficient and yet retains a biophysical interpretation, in virtue of being based on established neural mass models that are equipped with a slow dynamic on the parameters (such as synaptic rate constants or effective connectivity). In brief, we use an adiabatic approximation to summarise fast fluctuations in hidden neuronal states (and their expression in sensors) in terms of their second order statistics; namely, their complex cross spectra. This allows one to specify and compare models of slowly changing parameters (using Bayesian model reduction) that generate a sequence of empirical cross spectra of electrophysiological recordings. Crucially, we use the slow fluctuations in the spectral power of neuronal activity as empirical priors on changes in synaptic parameters. This introduces a circular causality, in which synaptic parameters underwrite fast neuronal activity that, in turn, induces activity-dependent plasticity in synaptic parameters. In this foundational paper, we describe the underlying model, establish its face validity using simulations and provide an illustrative application to a chemoconvulsant animal model of seizure activity.
... We perceive this change within the context of brain maturation, as previously shown for the classical cortical rhythms of sleep EEG in early brain development. 21,32,38,39 The duration and amplitude of scalp HFO remained unaffected by age in our study, in contrast to spikes that reportedly feature higher amplitudes and shorter duration in younger children. 32 This finding is intriguing, since we would expect the changes in skull thickness and conductivity with advancing age 40 to result in higher signal amplitude attenuation in older children. ...
High-frequency oscillations in scalp EEG are promising non-invasive biomarkers of epileptogenicity. However, it is unclear how high-frequency oscillations are impacted by age in the paediatric population. We prospectively recorded whole-night scalp EEG in 30 children and adolescents with focal or generalized epilepsy. We used an automated and clinically validated high-frequency oscillation detector to determine ripple rates (80-250 Hz) in bipolar channels. Children < 7 years had higher high-frequency oscillation rates (P = 0.021) when compared with older children. The median test-retest reliability of high-frequency oscillation rates reached 100% (iqr 50) for a data interval duration of 10 min. Scalp high-frequency oscillation frequency decreased with age (r = -0.558, P = 0.002), whereas scalp high-frequency oscillation duration and amplitude were unaffected. The signal-to-noise ratio improved with age (r = 0.37, P = 0.048), and the background ripple band activity decreased with age (r = -0.463, P = 0.011). We characterize the relationship of scalp high-frequency oscillation features and age in paediatric patients. EEG intervals of
≥
10 min duration are required for reliable measurements of high-frequency oscillation rates. This study is a further step towards establishing scalp high-frequency oscillations as a valid epileptogenicity biomarker in this vulnerable age group.
... Analogically, epileptic discharges are likely to cause abnormal brain network wiring and dynamics. Therefore, exploring the differences in topological features between epileptic and healthy brain networks has become a common method to detect epilepsy, but the detection effect is not always satisfactory (De Lathauwer et al., 2000;Booth, 2005;Subramaniyam and Hyttinen, 2013;Preti et al., 2014;Najm, 2018;Park et al., 2018;Rosch et al., 2018;Li and Cao, 2019). ...
Electroencephalograph (EEG) plays a significant role in the diagnostics process of epilepsy, but the detection rate is unsatisfactory when the length of interictal EEG signals is relatively short. Although the deliberate attacking theories for undirected brain network based on node removal method can extract potential network features, the node removal method fails to sufficiently consider the directionality of brain electrical activities. To solve the problems above, this study proposes a feature tensor-based epileptic detection method of directed brain networks. First, a directed functional brain network is constructed by calculating the transfer entropy of EEG signals between different electrodes. Second, the edge removal method is used to imitate the disruptions of brain connectivity, which may be related to the disorder of brain diseases, to obtain a sequence of residual networks. After that, topological features of these residual networks are extracted based on graph theory for constructing a five-way feature tensor. To exploit the inherent interactions among multiple modes of the feature tensor, this study uses the Tucker decomposition method to get a core tensor which is finally reshaped into a vector and input into the support vectors machine (SVM) classifier. Experiment results suggest that the proposed method has better epileptic screening performance for short-term interictal EEG data.
... Common approaches include tracking certain network measures over time (Sizemore & Bassett, 2018), using hidden Markov models (Eavani et al., 2013;Sourty et al., 2016;Vidaurre et al., 2018), and considering dynamic networks as multilayer networks (De Domenico et al., 2013;Kivelä et al., 2014;Sizemore & Bassett, 2018). Other recent approaches have used distance matrices to evaluate dFNs from fMRI (Cabral et al., 2017) and dynamic correlation matrices from scalp EEG (Rosch, Baldeweg, Moeller, & Baier, 2018). Future work should compare these different methods to our framework to find which one better characterizes dFNs in epilepsy and other contexts, and assess whether these approaches complement each other. ...
Evidence suggests that brain network dynamics are a key determinant of brain function and dysfunction. Here we propose a new framework to assess the dynamics of brain networks based on recurrence analysis. Our framework uses recurrence plots and recurrence quantification analysis to characterize dynamic networks. For resting-state magnetoencephalographic dynamic functional networks (dFNs), we have found that functional networks recur more quickly in people with epilepsy than healthy controls. This suggests that recurrence of dFNs may be used as a biomarker of epilepsy. For stereo electroencephalography data, we have found that dFNs involved in epileptic seizures emerge before seizure onset, and recurrence analysis allows us to detect seizures. We further observe distinct dFNs before and after seizures, which may inform neurostimulation strategies to prevent seizures. Our framework can also be used for understanding dFNs in brain function in health and other neurological disorders.
