Abstracts

Rationale: Recording of electroencephalograph (EEG) activity and video recording of animal behavior is necessary for the accurate assessment and quantification of seizure activity. In rodent models, visual analysis of seizure activity paired with EEG confirmation is the gold standard for determination of seizure events as a result of kindling models, genetic alteration, or drug effects. To help streamline the collection, identification and analysis of seizure events, we have developed an integrated software toolset for scoring and analysis of EEG-based seizure data and associated video records. Methods: Five mice (adult male wild type C57BL/6J) were injected with pilocarpine to induce status epilepticus (Pilocarpine model of epilepsy in mice). Five days after the pilocarpine-induced status epilepticus, mice were continuously monitored with video-EEG (Pinnacle Technology Model 8200) for at least 2 months to monitor the occurrence of spontaneous recurrent seizures. The data samples used for preliminary testing were 24 hours long and were taken 30 days after the status epilepticus. Rapid seizure identification within large datasets is accomplished using user-defined tools based on parameters such as seizure length, specific frequency, root mean squared (RMS) power and line length. Identified seizure events can further be classified using embedded video and rated using standard Racine Scale parameters. Individual seizure events can be isolated and exported as .avi video clips for use in presentations and paper submission. The software accepts input from a variety of data formats including .pvfs and .edf while analyzed data can be outputted to formatted ASCII text files for use with 3rd party statistical and spreadsheet programs. Results: The datasets were manually scored by an expert, and the data were then automatically scored by two different methods using the automated seizure analysis toolkit. Method 1: RMS power in the 1 to 100 Hz band. The results are presented as Seizures automatically found / Seizures found by the expert scorer. Mouse A: 3/3, Mouse B: 5/5 + 1 false positive, Mouse C: 3/3, Mouse D: 5/5 + 1 false positive, Mouse E: 5/5. Method 2: Line Length; Mouse A: 3/3, Mouse B: 5/5, Mouse C: 3/3, Mouse D: 5/5, Mouse E: 5/5. In these preliminary studies, some amount of manual threshold adjustment was necessary. In the next several months, the thresholds will be chosen for each dataset either automatically or via a training algorithm. Conclusions: Use of this type of integrated toolkit can aid with rapid scoring and analysis of large volumes of EEG-based seizure data.