Abstract

METHODS FOR THE DETECTION OF SEIZURE BURSTS IN EPILEPSY

Vora J. M.* and Karkar M. A.

ABSTRACT

Aim: Methods for the Detection of Seizure Bursts in Epilepsy. Background: Seizure clusters and “bursts” are of clinical importance. Clusters are reported to be a marker of antiepileptic drug resistance. Additionally, seizure clustering has been found to be associated with increased morbidity and mortality. However, there are no statistical methods described in the literature to delineate bursting phenomenon in epileptic seizures. Materials and Methods: We present three automatic burst detection methods referred to as precision constrained grouping (PCG), burst duration constrained grouping (BCG), and inter- seizure interval constrained grouping (ICG). Concordance correlation coefficients were used to confirm the pair-wise agreement between common bursts isolated using these three automatic burst detection procedures. Additionally, three graphical methods were employed to demonstrate seizure bursts: modified scatter plots, staircase plots, and drop-line plots. Burst detection procedures are demonstrated on data from continuous intracranial ambulatory EEG monitoring in a patient diagnosed with drug-refractory focal epilepsy. Results and discussion: We analyzed 1,569 seizures, from our assigned index patient, captured on ambulatory intracranial EEG monitoring. A total of 31, 32, and 32 seizure bursts were detected by the three quantitative methods (BCG, ICG, and PCG), respectively. The concordance correlation coefficient was ≥0.99 signifying considerably stronger than chance burst detector agreements with one another. Conclusion(s): Bursting is a quantifiable temporal phenomenon in epilepsy and seizure bursts can be reliably detected using our methodology.

Keywords: EEG, Bursts, Seizure cluster, Epilepsy, Inter seizure interval.