Automatic Seizure Detection in Rats using ECoG: A Comparison of Methods

Introduction: Precise automated seizure detection is a challenging problem. Nevertheless, the accurate detection of seizures plays an important role, especially in promoting the development of "closed-loop" therapies. Several detection methods have been developed and published so far. Mostly, the performance of these systems cannot be analysed or compared properly due to the use of heterogeneous data, varying seizure definitions or just because concrete information about the implementations are not available. Methods: Several new approaches for automatic seizure detection were implemented next to the most common methods available from the literature. Visually analysed ECoG obtained from rats with focal neocortical epilepsy upon local tetanus toxin plus cobalt chloride application have been used to compare the methods' performance. Results: None of the investigated methods was able to detect all existent seizures with reasonable degree of accuracy. Some methods showed a high performance for certain specific seizure types, but there is no "one size fits all" detection method. The implemented tool is suitable for performance tests and comparison of detection methods on arbitrary EEG/ECoG data. Conclusion: In the future, the present implementation can be used to evaluate and compare new seizure detection methods, and thus the performance of these can be verified fast and independently.