Epileptiform Spike Detection in Electroencephalographic Recordings of Epilepsy Animal Models Using Variable Threshold

Abstract

Epilepsy is a public health issue worldwide, given its biological, social, and economic impacts. Moreover, and particularly important, a significant portion of patients is refractory to conventional treatments and novel treatments are in need. By this token, the use and development of computational tools for the detection of epileptiform spikes, together with its feature extraction, have central significance, since these are recognized electrographic signatures of the disorder. In the present work, a detection method of such paroxysms in electroencephalographic recordings is proposed. With low mathematical complexity, the algorithm was developed for fast spike detection by using amplitude and time thresholds - both of them adjustable by the user - and applying a moving and variable amplitude threshold, calculated in each temporal window of analysis. This was done in order to provide greater adaptability to the algorithm and cope with the variable nature of epileptiform spikes. The algorithm was applied to recordings of animals submitted to acute seizures induced by a chemoconvulsant and results were compared to the visual detection of a specialist. Results showed the proposed algorithm can perform at the same level of other previously described approaches, considering the highly variable amplitude of spikes.