Abstracts

Rationale: Interictal spiking has been linked to impaired cognition. However, these electrographic events vary morphologically within and across patients, and the significance of this variance to cognition has yet to be systematically investigated. We hypothesize that morphologically distinct spike events differentially impact cognitive processes. Uncovering and understanding this differential impact could guide the development of future therapies. Methods: We detect interictal spikes in electrocorticograms using a previously validated template-matching algorithm (Horak et al., SPIE Optical Engineering+ Applications, 2015). We apply the multilevel one-dimensional discrete wavelet transform to each event, which yields a set of wavelet coefficients. We then use the mean-shift clustering algorithm to sort the events into morphologically similar groups. Finally, we calculate spiking rates for each cluster and fit a generalized linear mixed model (GLMM) to determine the relationship between spiking and performance on the task.  Results: We first validated the approach on a simulated data set and successfully produced realistic clusters of related events. We then analyzed data collected from 29 intracranial monitoring patients, all of whom completed several sessions of a free-recall memory task. Our approach identified distinct spike morphologies for each patient, only a subset of which correlated (p < 0.05) with performance on the task. Conclusions: Previous work from our group determined that rates of interictal spiking during the encoding (p < 0.01) and recall (p < 0.001) periods of the free-recall tast negatively correlated with memory performance (Horak et al., Epilepsia, 2017). Our latest findings expand this work by identifying morphologically distinct groups of spikes that differentially relate to task performance. Our results suggest that morphology should be considered alongside factors like location and spread in categorizing the cognitive impact of interictal spikes.  Funding: Defense Advanced Research Projects Agency (DARPA) Restoring Active Memory (RAM) program (Cooperative Agreement N66001-14-2-4032)