Abstracts

Rationale: The majority of intracranial EEG (iEEG) datasets are acquired in an acute setting while in the Epilepsy Monitoring Unit over a relative short period of time (1-14 days). This does not necessarily represent a patient’s neurophysiologic baseline. The University of Melbourne recently recorded long term continuous Intracranial Electroencephalogram (iEEG) recordings (>1 year) that will provide unique and helpful insight into our understanding, diagnosis, and treatment of epilepsy (Cook et al. 2013). We present here an overview of the approach and methods used to analyze such a dense and rich dataset looking specifically at the dynamics of the raw iEEG recording and time windowed features in the months after implantation.Methods: The dataset consists of fourteen patients implanted with two 8 channel electrode arrays placed over the determined epileptogenic zone and continuously recorded from for one to two years. The data from these patients were stored on IEEG.org, a cloud based research tool for sharing data, tools, and findings. A pipeline for analysis was established using Matlab that utilized parallel processing techniques in order to improve performance. Energy and line length features were calculated using a 15 second moving window over the first 60 days post-implantation for each patient. The feature average and standard deviation were then calculated for each day across all patients.Results: The analysis showed that energy and line length of a signal decreases over the first month and reaches a point of relative stability during the second month (Figures 1 & 2). Both energy and line length follow this trend, although there exists variability between each subject. This trend is typically not visible in data recorded from Epilepsy patients as the typical recording duration is significantly shorter.Conclusions: The preliminary results presented evidence of substantial feature variation in the iEEG signal over the first two months after implantation. This analysis could have significant clinical implications by providing insight into how iEEG signals are affected by the trauma and resulting tissue reaction that results from clinical pre-surgical evaluation as well as neuromodulators requiring electrode implantation. Further work is necessary to determine the robustness of these findings across channels and across various amplitude and spectral features.