Does sleep deprivation alter functional EEG networks in children with focal epilepsy?

Abstract

Electroencephalography (EEG) recordings after sleep deprivation increase the diagnostic yield in patients suspected of epilepsy if the routine EEG remains inconclusive. Sleep deprivation is associated with increased interictal EEG abnormalities in patients with epilepsy, but the exact mechanism is unknown. In this feasibility study, we used a network analytical approach to provide novel insights into this clinical observation. The aim was to characterize the effect of sleep deprivation on the interictal functional network organization using a unique dataset of paired routine and sleep deprivation recordings in patients and controls. We included 21 children referred to the first seizure clinic of our center with suspected new onset focal epilepsy in whom a routine interictal and a sleep deprivation EEG (SD-EEG) were performed. Seventeen children, in whom the diagnosis of epilepsy was excluded, served as controls. For both time points weighted functional networks were constructed based on interictal artifact free time-series. Routine and sleep deprivation networks were characterized at different frequency bands using minimum spanning tree (MST) measures (leaf number and diameter) and classical measures of integration (path length) and segregation (clustering coefficient). A significant interaction was found for leaf number and diameter between patients and controls after sleep deprivation: patients showed a shift toward a more path-like MST network whereas controls showed a shift toward a more star-like MST network. This shift in network organization after sleep deprivation in patients is in accordance with previous studies showing a more regular network organization in the ictal state and might relate to the increased epileptiform abnormalities found in patients after sleep deprivation. Larger studies are needed to verify these results. Finally, MST measures were more sensitive in detecting network changes as compared to the classical measures of integration and segregation.

Keywords: EEG; epilepsy; graph theory; minimum spanning tree; network analysis; sleep deprivation.

Figure 1

Two schematic illustrations of networks. (A) a standard network and (B) a MST network wherein all nodes are connected only once resulting in a loopless network. In panel (B) the black lines represent the most efficient connections in the MST network; grey lines represent the excluded functional connections.

Figure 2

Three network topologies based on MST network. On the left a path-like topology with few leafs and long diameter; on the right a star-like topology with many leaves and a moderate diameter. In the middle an intermediate form combining the qualities of a line-like and star-like topology. Notify that all networks have the same number of nodes and connections. Leafs colored in green.

Figure 3

Illustration of interaction effects from MST measures diameter (left graph) and leaf number (right graph) per frequency band as revealed with a repeated measures ANOVA (mean values and standard error of the mean bars). In this analysis we included only patients in whom the SD-EEG was of added value (n = 15), and all controls (n = 17) (Table 4). There was a significant interaction for diameter in the alpha band; the diameter increased in patients whereas an opposite effect was found for controls. For leaf number, a significant interaction was found in the alpha band; the leaf number decreased in patients whereas an opposite effect was found for controls. Together, these results in the alpha band suggest a shift toward a path-like topology for patients after sleep deprivation and a shift toward a star-like topology for controls. ^*Significant (p < 0.05), ^**Significant after correcting for multiple post-hoc comparisons (false discovery rate test).