Abstracts

Rationale: Seizures are characterized by abnormal, synchronous, and excessive neuronal activity, which may result in loss of consciousness. Prior studies demonstrated correlations between consciousness and loss of scalp EEG signal complexity in patients emerging from anesthesia. Multiscale Entropy (MSE) and Correlation Dimension (CDim) can be utilized to quantify the complexity of biological signals, including EEG. Scalp EEG is readily used for clinical and experimental studies. However, this modality is susceptible to noise and has limited spectral bandwidth. In contrast, intracranial EEG is rarely contaminated with noise and captures a wider bandwidth, especially higher frequencies. The purpose of this study was to quantify the complexity of intracranial EEG during seizures as compared to awake and sleep states in humans.


Methods: Intracranial depth electrodes were placed using stereotactic image-guided approach to localize epileptic foci in 7 patients with medically refractory epilepsy. Electrode targeting was determined by clinical factors such as seizure semiology and imaging findings. EEG was sampled at 1000 Hz. Recordings were segmented into multiple periods of awake, sleep, and seizure states. After randomly selecting a 20-second clip from each of these states, MSE at 6 different time scales (1, 11, 21, 31, 41, and 51 samples) and CDim (dimension of 2 and lag of 1) were calculated. One-way ANOVAs were performed on these 7 complexity measures to determine any statistical difference (p < 0.05) between all the states. Post-hoc pairwise Tukey–Kramer tests were then used to assess statistical differences between the individual states.


Results: Overall, the complexity analyses revealed the lowest complexity during seizure, followed by sleep (intermediate) and awake (highest) (Figure 1). Furthermore, with MSE, larger timescales were associated with higher complexity. There was a statistically significant difference (p < 0.05) in complexity among all the states with the one-way ANOVA analysis for MSE at time scales 11, 21, 31, 41, and 51, but not at time scale 1 or with the CDim (MSE1: p = 0.098; MSE11: p = 0.001; MSE21: p = 0.003; MSE31: p = 0.008; MSE41: p = 0.014; MSE51: p = 0.018; CDim = 0.121) (Table 1). Utilizing the Tukey-Kramer pairwise comparisons, there was a statistically significant decrease in complexity between awake compared to seizure for MSE with all time scales except with time scale 1 and not with CDim (MSE1: p = 0.088, MSE11: p = 0.001, MSE21: p = 0.003, MSE31: p = 0.010, MSE41: p = 0.010, MSE51: p = 0.014, CDim = 0.106). There were no significant differences in complexity measures between awake and sleep except with MSE11 (p = 0.011). In addition, there were no significant differences in complexity measures between sleep and seizure.


Conclusions: Our findings demonstrate a significant decrease in intracranial EEG complexity during seizures compared to awake state. In contrast, there was little to no statistical difference in signal complexity between comparisons of other states. Future studies exploring other aspects of intracranial recordings will further understanding of seizures and consciousness.


Funding: None