Abstracts

Detrended Fluctuation Analysis (Peng, CK et al. Phy Rev E 1994;49 1685-1689) has proven to be a robust method for characterizing long-range temporal correlations in signals from complex non-stationary systems. This algorithm has revealed power-law scaling behavior when applied to scalp electroencephalography and magnetoencephalography. We applied DFA to intracranial EEG signals with respect to sleep, wakefulness, preictal state and location in epileptogenic cortex. 146 iEEG 20 minute recordings were analyzed from 684 intracranial cortical electrodes in 6 patients with medical refractory epilepsy. Sleep and wake recordings were selected based on video inspection and presence or absence of sleep related EEG changes. Pre-ictal recordings were selected prior to EEG changes suggestive of ictal onset. DFA scaling [alpha] exponents were computed in 5-10 sleep vs wake states, and 3-10 pre-ictal states within each patient. State dependent relationships in DFA exponents of each recording site were then determined and correlated with location in epileptogenic or nonepileptogenic cortex, using unpaired t-test and anova statistical methods with statistical signficance reported only for p[lt]0.01. DFA scaling exponents from analysis of all recordings showed power-law behavior (0.5[lt][italic][alpha][lt]1)[/italic]. Significant differences in mean [italic][alpha] [/italic]exponents were found between pre-ictal, wake, and sleep states. These respective differences varied between patients and as a function of location in epileptogenic or nonepileptogenic cortex. Significant differences between sleep and wake [italic][alpha][/italic] were seen in all 6 patients at epileptogenic sites, and 5 of 6 patients at nonepileptogenic sites. Pre-ictal [italic][alpha][/italic] in nonepileptogenic sites were significantly different from wake in all patients and in epileptogenic sites in 5 of 6 patients. Pre-ictal [italic][alpha][/italic] differed from sleep in all patients at epileptogenic sites but in only 3 of 6 patients at nonepileptogenic sites. This study of DFA on cortical iEEG demonstrates the presence of long-range power law scaling behavior with distinct state dependent differences in epileptogenic and nonepileptogenic cortex. DFA differences between pre-ictal vs sleep and wake recordings were found in all 6 patients but were most significant in nonepileptogenic cortex for preictal vs wake states and in epileptogenic cortex for preictal vs sleep states. These findings may have important implications in seizure prediction and clinical epilepsy. (Supported by NIH R01-MH55687, NIH R01-MH061975
Brigham and Women[apos]s Hospital training grant for translational neuroscience.)