Neural Connectivity in Epilepsy as Measured by Granger Causality

Abstract

Epilepsy is a chronic neurological disorder characterized by repeated seizures or excessive electrical discharges in a group of brain cells. Prevalence rates include about 50 million people worldwide and 10% of all people have at least one seizure at one time in their lives. Connectivity models of epilepsy serve to provide a deeper understanding of the processes that control and regulate seizure activity. These models have received initial support and have included measures of EEG, MEG, and MRI connectivity. Preliminary findings have shown regions of increased connectivity in the immediate regions surrounding the seizure foci and associated low connectivity in nearby regions and pathways. There is also early evidence to suggest that these patterns change during ictal events and that these changes may even by related to the occurrence or triggering of seizure events. We present data showing how Granger causality can be used with EEG data to measure connectivity across brain regions involved in ictal events and their resolution. We have provided two case examples as a demonstration of how to obtain and interpret such data. EEG data of ictal events are processed, converted to independent components and their dipole localizations, and these are used to measure causality and connectivity between these locations. Both examples have shown hypercoupling near the seizure foci and low causality across nearby and associated neuronal pathways. This technique also allows us to track how these measures change over time and during the ictal and post-ictal periods. Areas for further research into this technique, its application to epilepsy, and the formation of more effective therapeutic interventions are recommended.

Keywords: Granger causality; connectivity; connectivity analysis; epilepsy; seizures.

Figure 1

EEG (longitudinal/sequential montage) sample showing focal (T5, P3) spike and wave discharges for case 1.

Figure 2

EEG (longitudinal/sequential montage) sample showing focal (F4, T4) spike and wave discharges for case 2.

Figure 3

SIFT/Granger (GGC) causality sequence of processing.

Figure 4

SIFT/Granger causality baseline image (case 1) for eight component dipoles localized to the right superior frontal gyrus (Broadman 6), left superior occipital gyrus, right medial frontal gyrus white matter, left thalamus, left middle frontal gyrus white matter, left precentral gyrus, right superior frontal gyrus (Broadman 9), and the left middle temporal gyrus. Smaller, dark blue circles and dark blue, thin lines indicate lower causality. This causality image at baseline (pre-ictal) occurs at −0.156 s.

Figure 5

SIFT/Granger causality baseline image (case 1) for 8 component dipoles and how they change over time. The sequence goes from 0.089 s (top left) to 0.195 s (top right) to 0.42 s (bottom left) to 1.72 s (bottom right).

Figure 6

SIFT/Granger causality baseline image (case 2) for nine component dipoles localized to the right anterior cingulate gyrus (Broadman 32), right superior temporal gyrus (Broadman 22), right precuneus white matter, left postcentral gyrus (Broadman 43), right medial frontal gyrus white matter, left middle frontal gyrus (Broadman 46), left precuneus (Broadman 39), left supramarginal gyrus (Broadman 39) and the left middle frontal gyrus white matter. Smaller, dark blue circles and dark blue, thin lines indicate lower causality. This causality image at baseline (pre-ictal) occurs at −0.0391 s.

Figure 7

SIFT/Granger causality baseline image (case 2) for nine component dipoles and how they change over time. The sequence goes from 0.313 s (top left) to 0.43 s (top middle) to 0.781 s (top right) to 0.898 s (bottom left) to 1.25 s (bottom middle) to 1.72 s (bottom right).