Abstracts

Rationale: The relationships between interictal spikes and seizures are of fundamental importance in clinical epilepsy, but remain poorly understood in many respects. In the focal syndromes, it is unclear which features of EEG spikes are relevant to epileptogenic potential (EP: the propensity for clinical seizures). We investigated whether the degree of morphological variability (MV) of spikes, under standard recording conditions, was correlated with EP. If MV is due to variable activation of different portions of an epileptogenic network, we hypothesized that increased MV also reflected a greater tendency for runaway change (i.e., seizures). Formally, this borrowed from the concept of ‘criticality’ in physical systems (e.g. Bak, P., 1996. How Nature Works. New York: Springer-Verlag) that undergo frequent irregular fluctuations and occasional large disturbance.Methods: We studied three archetypal, and clinically opposed, situations: periodic lateralized epileptiform discharges (PLEDs), comprising high-amplitude repetitive spikes with high EP, the spikes of benign focal epilepsy of childhood (BFEC), often equally outstanding in a typical EEG but of low EP, and the generalized periodic pattern (GPP) of severe encephalopathy, not normally considered possessing EP, though visually ‘epileptiform’. Digital EEG from routine 20-minute recordings were analyzed from three groups of five patients each with these diagnoses. A single representative channel from each multichannel trace was processed. Spikes were detected and clustered into morphological subgroups with an adapted version of WAVE_CLUS (Quian Quiroga et. al., 2004. Neur Comput 16:1661-87; software freely downloadable at www.vis.caltech.edu/~rodri). The software identifies spikes by amplitude thresholding followed by projection onto a wavelet basis. Morphological differences in spike populations translate to differences in the wavelet coefficients, which are used for reconstruction and clustering. The clustered and reconstructed data were compared across the three diagnostic classes.Results: PLEDs were the most morphologically variable, with multiple waveform types identified by WAVE_CLUS over the available range of parameter values. BFEC spikes were less so, with comparatively fewer subtypes identified over the same parameter range. The complexes of GPP were the least. Conclusions: WAVE_CLUS, a publicly-available software tool designed to classify spikes in experimental single-unit preparations is readily modified for use in routine clinical recordings. Information regarding epileptogenicity is implicit in spikes extracted from scalp EEG. MV of PLEDs, BFEC discharges and GPP relate to EP. This may represent the differing excitable properties of the underlying neural generators. The results illuminate clinical rules-of-thumb (e.g. ‘PLEDs-plus’ are more epileptogenic than PLEDs (Reiher et. al., 1991. Electroencephalogr Clin Neurophysiol 78:12-7)) and may inform controversies in clinical neurophysiology (Chong & Hirsch, 2005. J Clin Neurophysiol 22:79-91), while aiding conceptualization of epileptogenesis.