We present a method for automatic detection of seizures in EEG that might help clinicians by speeding up the process of seizure detection. The method consists of extraction of Log-Energy Entropy from band-passed EEG and use of a Support Vector Machine (SVM) classifier. Furthermore, using multiple regression analysis, we evaluated the effect of some characteristics of the patients on the performance of the method. We found that the type of epilepsy is the major factor, which influenced the performance of the method. The high performance of the method makes it feasible also for real-time applications.