Identifying the epileptic network by linking interictal functional and structural connectivity

Abstract

Over the last two decades, it has become increasingly clear that epilepsy is a network disorder. However, it is unclear whether these networks are established only during seizures or persist interictally. The goal of this study was to identify whether functional seizure networks exist interictally and evaluate if there is a structural basis to these networks. We identified four patients with mesial temporal lobe epilepsy who underwent resective epilepsy surgery. We estimated functional and structural connectivity across intracranial electrode contacts involved in seizure onset, early spread, and uninvolved controls. Across all interictal epochs considered, we found higher functional and white matter connectivity across cortical regions involved in seizure spread. Additionally, we observed that the patient in our cohort with the best seizure outcome had the highest functional connectivity across seizure contacts. Functional connectivity findings suggest the presence of an interictal seizure network that parallels underlying structural connectivity. Furthermore, our findings suggest that disruption or ablation of highly connected seizure regions may be necessary to achieve improved post-operative seizure freedom.

Keywords: Diffusion tensor imaging; Epilepsy; Functional connectivity; Interictal connectivity; Mesial temporal lobe epilepsy; Seizure networks.

Fig. 1

Experimental workflow. (A) Here, we show a schematic of the experimental workflow. SOZ and SP contacts were identified by two epileptologists (P.F. and J.P.) through examination of seizure recordings during intracranial neuromonitoring. Equidistant control contacts were identified, with the aim of choosing cortical areas covered by subdural electrodes that were approximately equidistant from the SOZ. Subsequent analysis involved extraction of interictal (at least 6 h from any seizure event) ECoG and comparison of MSC between SOZ-SP and SOZ-control contact pairs. (B) Here, we show a representative MSC plot for a single pair of SOZ-SP (red) and SOZ-control (blue) observations. Abbreviations: ECoG - electrocorticography, SOZ - seizure onset zone, SP - primary seizure spread zone, MSC (γ) - magnitude-squared coherence.

Fig. 2

Increased interictal functional connectivity. Here, we show plots of coherence-based FC using MSC in the 1–30 Hz spectral band. (A) Here, we show that pooled MSC values are significantly higher for all SOZ-SP pairs compared to SOZ-control pairs (Wilcoxon matched-pairs signed rank test, p = 0.0240). In other words, there is higher FC interictally between cortical regions involved in seizure activity compared to equidistant control regions, suggesting the presence of a pathologic epileptic network that persists interictally. (B) Here, we took the ratio of MSC of SOZ-SP with respect to SOZ-control, to normalize for within patient variability. Each data point refers to the coherence ratio computed from a single interictal ECoG recording from each patient. Specifically, we show that the patient with the best post-operative seizure outcome at follow-up (Patient 1, ILAE = 1) had significantly higher MSC ratios compared to those with poor outcome (Patients 2–4, ILAE ≥ 2) (F(3,16) = 10.91, p = 0.0004, ordinary one-way ANOVA with corrections for multiple comparisons). Abbreviations: MSC - magnitude-squared coherence, FC – functional connectivity, SOZ - seizure onset zone, SP - primary seizure spread zone, ECoG – electrocorticography.

Fig. 3

Higher structural connectivity among seizure regions. (A) Here, we obtained white matter tract connectivity estimates across all parcel pairs of interest (i.e., SOZ, SP, and control). We show that there is significantly increased white matter tract connectivity across cortical regions involved in seizure activity (SOZ-SP) compared to uninvolved (SOZ-control) (Wilcoxon matched-pairs signed rank test, p = 0.0073). (B) Here, we fit a model to time to involvement and white matter tract connectivity data to see if white matter tract connectivity could be informative of seizure spread time. Specifically, we fit a linear model on log transformed time to involvement and white matter tract connectivity values, given that the data appeared to follow a logarithmic fit. We show that slope of the line of best fit of the log-transformed data is significantly nonzero (Y = − 0.6583∗X + 0.7611, p = 0.0008). In other words, higher white matter tract connectivity is significantly negatively correlated with seizure spread time, suggesting a role for structural connectivity patterns in seizure spread and the utility of DTI in elucidating the epileptic network in patients implanted with cortical grid and depth electrodes. Abbreviations: SOZ - seizure onset zone, SP - primary seizure spread zone.