Investigation into the Dynamic Repertoire and Functional Connectivity in Focal Epilepsy
Poster No:
1709
Submission Type:
Abstract Submission
Authors:
Oliver Sherwood1, David Carmichael2, Joel Winston3, Robert Leech2
Institutions:
1Kings College London, London, United Kingdom , 2King's College London, London, United Kingdom, 3Kings College London, London, United Kingdom
First Author:
Co-Author(s):
Introduction:
Epilepsy is a neurological disorder characterised by abnormally coordinated brain activity which leads to seizures. Treatment for epilepsy is managed with anti-seizure medication that is effective in 70% of patients; however, there is still a large proportion of people unable to achieve seizure freedom [1]. Our current understanding of epilepsy is as a network disorder, largely due to imaging findings that demonstrate focal and generalised epilepsy being characterised by macroscopic network perturbations. However, an improved understanding of network dynamics and their perturbation in epilepsy could provide new diagnostic and therapeutic avenues [2]. This study aims to investigate the dynamic repertoire of patients with focal epilepsy (FE) using Leading Eigenvector Dynamic Analysis (LEiDA) to describe functional network dynamics in BOLD (Blood Oxygen Level Dependent) fMRI [3]. Specifically, we aimed to determine the occurrence and stability of FC brain states in a group of FE patients compared to healthy controls (HC).
Methods:
This study used 55 participants consisting of patients with Focal Epilepsy (35 drug-resistant children aged 7-18 years) and healthy controls (20 children without Epilepsy aged 7-17 years). First, BOLD phase coherency connectivity is used to obtain a time-resolved dynamic Functional Connectivity (dFC) matrix for each run. A Hilbert transform is used to calculate phase of the acquired BOLD signal, and the phase coherence is calculated at each time point (t) between each of the 90 brain regions (n, p) of the AAL atlas parcellation. We consider only the leading eigenvector of each dFC(t). We next apply k means clustering analysis of all the leading eigenvectors across time-points and subjects (varying the k values between 2 and 20 clusters and obtain k centroid clusters each representing an Nx1 vector of the recurrent FC state). We next compare our obtained FC states to 7 a-priori resting state networks (RSN's) defined by Yeo et al [4] as well as delving into the FC state trajectory amongst the participants (achieved by calculating: mean fractional occupancy, dwell time and the transition probability of the FC states). Statistical validity was assessed using a Monte Carlo Permutation test which functions by calculating a test statistic (in this case Levene's t-test) for the obtained group assignments, before reshuffling the data and thus group assignments and recalculating the test statistic.
Results:
Following analysis, the optimal clustering was obtained using the Dunn index at k = 13 (figure 1). Fractional occupancy of brain states resembling limbic and visual network were significantly different between the two groups (p<0.05). In dwell time, a significant difference between groups was only seen in one brain state (brain state = 9, visual network) (p<0.05) (figure 2). The transition matrix showed a significant difference in transitions from FC state 3 (visual) to FC state 9 (visual) as well as from FC state 8 (frontal parietal) to FC state 13 (unlabelled).
Conclusions:
From these results it appears that FE patients are less likely to engage the limbic network than HCs but more likely to dwell in the visual network. It is interesting to note that there is no observed propensity for FE patients to reside within the DMN as has been observed in other research [5]. While it is imprudent to draw significant inferences from these preliminary results, it is clear that this method of analysis may prove vital in unpicking the variations in functional repertoire and transitions in patients with epilepsy compared to healthy controls. Further improvement of the analysis methods, to overcome limitations and diversify the dataset are required to fully investigate this. Some of these limitations include reliance on the AAL parcellation, heterogeneity within the FE group and unwanted decomposition of important signals from the data.
Modeling and Analysis Methods:
Connectivity (eg. functional, effective, structural)
fMRI Connectivity and Network Modeling 1
Task-Independent and Resting-State Analysis 2
Other Methods
Novel Imaging Acquisition Methods:
BOLD fMRI
Keywords:
Epilepsy
Modeling
MRI
Statistical Methods
Other - LEiDA; Dynamic Functional Connectivity (dFC)
1|2Indicates the priority used for review
Provide references using author date format
[2] M. Centeno and D. W. Carmichael, ‘Network Connectivity in Epilepsy: Resting State fMRI and EEG-fMRI Contributions’, Front. Neurol., vol. 5, 2014.
[3] J. Cabral et al., ‘Cognitive performance in healthy older adults relates to spontaneous switching between states of functional connectivity during rest’, Sci. Reports 2017 71, vol. 7, no. 1, pp. 1–13, Jul. 2017.
[4] B. T. Thomas Yeo et al., ‘The organization of the human cerebral cortex estimated by intrinsic functional connectivity’, J. Neurophysiol., vol. 106, no. 3, pp. 1125–1165, Sep. 2011.
[5] N. B. Danielson, J. N. Guo, and H. Blumenfeld, ‘The default mode network and altered consciousness in epilepsy’, Behav. Neurol., vol. 24, no. 1, p. 55, Jan. 2011.