Abstracts

Rationale: The primary objective in the presurgical evaluation of epileptic patients is to identify the brain region(s) responsible for generating the patient’s habitual seizures. Visual inspection of the electroencephalogram (EEG) as well as imaging studies are often inconclusive. Automated epileptogenic focus localization by mathematical analysis of interictal scalp EEG holds a theoretical and practical potential to contribute to this challenge and therefore to a more accurate diagnosis and treatment of epilepsy. No such scalp EEG-based focus localization algorithm with adequate sensitivity and specificity has been reported previously, perhaps due to the large variety of epilepsy types, normal EEG variants, and different nonepileptiform focal abnormalities that occur. Despite these potential confounding factors, we have developed an algorithm that reliably accomplishes this goal with a high degree of sensitivity and specificity. Methods: Transfer Entropy (TE) is a recently proposed measure of the information flow between coupled linear or nonlinear systems. In the past, we have shown the successful application of an improved TE method on long (days), continuous, intracranial EEG from patients with focal temporal lobe epilepsy (TLE) for localization of their foci. Motivated by these results we extended our analysis to scalp EEG recorded from five patients with clinically well-defined focal TLE to first study the sensitivity of our focus localization algorithm. In addition, a set of 5 “control” EEGs recorded from nonepileptic seizure (NES) patients was also obtained to test the specificity of the algorithm. TE profiles per electrode site were estimated from successive, non-overlapping scalp EEG segments of 10.24 seconds in duration (2048 points per segment at 200 Hz sampling rate) for the first six hours of EEG recording per patient after admission to the epilepsy monitoring unit. Statistical metaprocessing of the TE values, was performed to study focus localization. Results: Based on statistical analysis of the TE profiles, the epileptogenic focus was successfully identified interictally in each of the five TLE patients as the electrode site or set of electrode sites that significantly drives the rest of the electrode sites (high sensitivity). The estimated foci were in agreement with the clinically assessed sites of the epileptogenic focus in all patients analyzed. In addition, the focus localization algorithm did not find any electrode as a possible driver electrode in the patients with NES (high specificity). Conclusions: These robust findings suggest that a) our proposed interictal focus localization algorithm is highly sensitive and specific and therefore, may prove to be a reliable clinical tool for prospective localization of the focus/foci by using short interictal scalp EEG segments, and b) non-epileptic events may be characterized by the absence of significant driver electrodes and therefore, may serve as a tool to test the specificity of any prospective real-time focus localization algorithm or even to clinically distinguish them from epileptic seizures.