Abstracts

The use of nonlinear analysis methods to detect patterns in EEG recordings that may predict impending seizures has been well documented. These methods may also prove useful in locating the seizure onset zone, a process which is usually performed by visual inspection of the EEG and which forms the basis for tailoring surgical resections to optimize epilepsy surgery outcomes.
An algorithm proposed by Quiroga, Kreuz, and Grassberger (Physical Review E 66,041904 2002) measures synchronization between timed events in a set of signals and their relative time-delay characteristics, where the definition of an event depends on the particular application. We used this algorithm to analyze intracranial EEG recordings of patients undergoing monitoring for seizure localization. During monitoring, the location and timing of seizure onset were determined based on clinical data and visual inspection of the ictal EEG. In our implementation, we defined an event as a minimum peak in a negative waveform relative to an inactive reference in the frequency band (usually gamma range) that was most prominent at seizure onset. We then calculated a synchronization measure for each pair of electrodes and their delay behavior.
We found that an increase in synchrony occurred at seizure onset, and that the electrodes at which leading synchronous events occur were located in the clinically-suspected seizure onset zone. As the seizure progressed, the identity of the leading electrodes evolved, but remained within the onset zone.
Our findings suggest that analysis of event synchronization and time delay patterns may be useful in the clinical determination of the seizure onset zone, and in tracking the evolution of the seizure focus during a clinical episode. We believe that further investigation of this and other techniques that detect asymmetries in synchronous events is warranted.
[Supported by: NICHD/NIBIB Mentored Biomedical Engineering Career Development Award]