Brain mapping of an epileptic brain using EEG functional and effective connectivity
-
1
University of Ghent, Data Analysis, Belgium
-
2
UCL Institute of Neurology, Department of Clinical and Experimental Epilepsy, United Kingdom
-
3
Laboratory for Clinical and Experimental Neurophysiology, Belgium
The topology on the network obtained from functional and effective connectivity in the epileptic brain can be of great importance for the detection of the transition from interictal to ictal states and the localization of the brain region where the seizure originates (Spencer, 2002). The methods that have been developed in this direction are able to explore brain dynamics across different brain regions, by exploiting the high temporal resolution of electroencephalographic intracranial and scalp recordings.
For this study, aimed to provide a review and comparative evaluation of connectivity methods applied to epilepsy, we consider data from two patients with refractory epilepsy. The dataset from the first patient consists of 27 scalp electrodes, 1 amygdalo-occipital depth electrode with 8 contact points, 4 grids with 4 contact points located on the left and right temporobasal area and 2 grids of 6 contact points located on the left and right temporolateral area. This patient manifested 6 seizures during the recording.
The second dataset consists of 27 scalp electrodes, 1 depth electrode with 8 contact points, 2 grids with 4 contact points on the right and left temporal area, 3 grids with 6 contact points on the right frontal, left frontal and left temporal areas. This patient manifested 3 seizures during the recording.
From both datasets we analysed segments of 120s before the electrographic seizure onset and a segment of equal length after the end of the seizure, as well as the ictal period itself.
We hypothesize that directed connectivity measures (such as Granger Causality or its analogues in frequency domain, Directed Transfer Function and Partial Directed Coherence) can shed light on the mechanisms/dynamics that occur during the ictal period and thus provide crucial tools both for focus localization and early seizure detection. Connectivity matrices are built for each data segment for all the proposed methods, corresponding to pre-ictal, ictal and post ictal states.
The computation of the connectivity matrices renders feasible the extraction of the total information flow from each node for all the measures (Coherence, DTF, PDC) but also the ingoing and outgoing information from each node in cases of the directed measures (DTF, PDC). The rank of the connectivity matrix indicates the number of the linearly independent rows or columns. Thus, tracking the rank of the connectivity matrices helps to detect the transition to a more organized state in brain activity and thus, gathering relevant information on the dynamics of the seizure onset. In this analysis we quantify the rank of the matrix my means of Singular Value Decomposition as described in Santaniello et al. (2011)
Indeed both looking at the information flow detected with these techniques or at the largest singular value of the corresponding connectivity matrices we find relevant indications on putative epileptogenic regions (significant differences in connectivity in pre-ictal and ictal states). More specifically, for the 27 scalp electrodes in both patients, coherence captured an increase in the maximum singular value around the time marked as intracranial electroencephalographic onset, while this trend continued until after the end of the seizure . High values of the maximum singular value indicate less diversity but stronger components which is in agreement with the concept that during the seizure the brain enters a more organized state (Iasemidis et al., 2004). For the cortical contacts of the strips and subdural grids of the two patients, a similar trend is detected compared to the one in the scalp electrodes, with an increased maximum singular value during the electroencephalographic onset for most of the used methods.
Being motivated to investigate why, how and which brain sources have caused the patterns that are being captured by our exploratory measures, we plan to incorporate in our work an application of a biologically inspired methodology, Dynamic causal Modelling (DCM).
There results obtained by the exploratory measures are going to be used in order to extract information for identifiable patterns and features of our datasets e.g. changes in connectivity in the different segments, identification of nodes as information sink or sources.
Exploiting this information will allow the individuation smaller networks of interconnected regions of interest on which DCM is going to be applied. The fact that DCM provides posterior estimates of neurobiological interpretable quantities (Stephan et al.,2010), can shed light not only on which neuronal populations cause these specific features and patterns in our data, but also how each subpopulation exerts over another (effective connectivity), which are the strengths of the synaptic connections among these subpopulations etc.
Keywords:
EEG dynamical connectivity,
neural networks,
Epileptic seizures,
focus localisation,
singular value decomposition
Conference:
Belgian Brain Council, Liège, Belgium, 27 Oct - 27 Oct, 2012.
Presentation Type:
Poster Presentation
Topic:
Higher Brain Functions in health and disease: cognition and memory
Citation:
Papadopoulou
M,
Leite
M,
Meurs
A,
Carrette
E,
Raedt
R,
Vonck
K,
Boon
P and
Marinazzo
D
(2012). Brain mapping of an epileptic brain using EEG functional and effective connectivity.
Conference Abstract:
Belgian Brain Council.
doi: 10.3389/conf.fnhum.2012.210.00064
Copyright:
The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers.
They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.
The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.
Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.
For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.
Received:
30 Aug 2012;
Published Online:
12 Sep 2012.
*
Correspondence:
Miss. Margarita Papadopoulou, University of Ghent, Data Analysis, Ghent, 9000, Belgium, marg.papadop85@gmail.com