Abstracts

RATIONALE: Extraction and subsequent analysis of time-series information form the foundation by which the behaviour of systems both physical and biological can be understood over time. An epileptic seizure may be considered as transient stabilization of quasi-periodic states within the dynamical repertoire of a population of hyper-synchronized neurons (Perez Velazquez et al., 1999). We present a simple method, Recursive Peak Time (RPT) analysis, based on peak detection, recursive plotting, and nonlinear mapping for time-series processing of electrophysiological recordings. We demonstrate RPT as a tool for visualization and quantification of electrophysiological signal changes relatable to quasi-periodicities. Analyses revealed characteristic signal changes that were interpreted to anticipate seizures.
METHODS: Hippocampal brain slice recordings were obtained from male Wistar rats (17-25 days old). Spontaneous seizure-like events (SLEs) were brought on by perfusing slices in 0.5 mM Mg+2. Single or multi-channel extracellular responses were recorded from CA1, CA3, and DG regions of the hippocampus. Intracranial EEG recordings of seizures were obtained from implanted depth electrodes in three patients, with unilateral (mesial) and bilateral temporal lobe seizure disorders, undergoing pre-surgical EEG monitoring. Continuous, interictal and ictal epochs of EEG were digitized at 200 Hz. Software was written to detect peaks and fast-transients using amplitude and width criteria. Recursive (return) maps were constructed from the interval data using appropriate delays. Resulting return plots were mapped onto an adjustable nonlinear surface that was used to unravel nontrivial temporal self-similarities in the time series. This allowed for visualization and quantification of activity over time.
RESULTS: RPT analysis of slice and EEG recordings revealed common and characteristic temporal trends that were used to anticipate seizures. In slice recordings signal changes were observed ~ 44 [plusminus] 33 s (average:stdev) in anticipation of the actual seizure-like event. Approximately 75 % of detected anticipatory events occurred within 50 s of SLE onset. In the case of EEG, this analysis detected electrographic changes ~ 29 [plusminus] 13 s in anticipation of electrographic onset. Approximately 75 % of detected anticipatory events occurred within 30 s of electrographic seizure onset.
CONCLUSIONS: We present RPT analysis as a concise method suitable for the analysis of electrophysiological recordings in the context of seizures. The analysis procedure and subsequent visualization tactics are able to extract meaningful information about temporal relations in time-series data. RPT is simple yet robust in quantitatively detecting electrographic signal changes, which make it useful for real-time implementation. This technique contributes to the investigation of dynamical mechanisms responsible for seizure onset by serving as a diagnostic tool, in conjunction with other available methodologies, for visualizing and quantifying electrophysiological signal changes. (1) Perez Velazquez et al. 1999, Eur. J. Neurosci. 11:2571-2576.
[Supported by: Funded by research grants from Citizens United for Research in Epilepsy (CURE) and the Savoy Epilepsy Foundation.]