Comment on: "Performance of a seizure warning algorithm based on the dynamics of intracranial EEG".

s With great interest we read the article of haovalitwongse et al. (2005) concerning the perormance of an automated seizure warning system ASWS) based on concepts from nonlinear dynamcs. To assess the performance of their algorithm, the uthors divided long-term intracranial EEG data from 0 patients into training and test data sets. For the trainng data, the authors reported a high prediction perforance with an average sensitivity of 76.12% accepting n average false prediction rate of 0.17 false warnings er hour. For the test data, an average sensitivity of 8.75% and an average false prediction rate of 0.15 ere obtained. These promising results nurture the hope to estabish a therapeutic device for epilepsy patients based n an in-time seizure warning. However, seizure preiction suffers from the intrinsic problem that high