Abstracts

Rationale: Prolonged video-EEG monitoring in chronic epilepsy rodent models has become an important tool in pre-clinical drug development of new therapies, in particular for anti-epileptogenesis, disease modification and drug resistant epilepsy. We have developed a user-friendly, easy to use, reliable, computational tool for detection of electrographic seizures from prolonged EEG recordings in rodent models of epilepsy. Methods: We applied a novel method based on advanced time-frequency analysis which detects the episodes of EEG with excessive activity in certain frequency bands. The method uses an original technique of short term spectral analysis - the Similar Basis Function algorithm. The algorithm may calculate the Fourier transform with arbitrary frequency resolution only within the frequency band of interest, which significantly reduces the number of necessary computations and increases the processing speed. The method was applied for off-line seizure detection from long-term EEG recordings from four spontaneously seizing, chronic epilepsy rat models: post-status epilepticus model of temporal lobe epilepsy (n=59 rats, n=214 seizures); the fluid percussion injury model of post-traumatic epilepsy (n=5 rats, n=49 seizures); and two genetic models of absence epilepsy – GAERS and WAG/Rij (n=41 and 14 rats, n=8733 and 825 seizures respectively). Results: High values of power spectrum in frequency band 17-25 Hz were found to specifically indicate seizures in all four animal models. This peculiarity comes from the frequency composition of individual spike-wave complexes (ISWC) within the seizures in these animals: our original comparative analysis revealed that the amplitude spectra of ISWCs of (at least) these four rat models are remarkably similar and have a single expressed peak within 17 – 25 Hz frequency range. Focusing on this band, our computer program detected 100% of seizures in all 119 rats. Electrode artefacts, which are usually present in long-term EEG recordings, may also significantly contribute to this frequency band, so they were also selected by the program. This selection, however, generated a very low rate of false positives. For their elimination, a quick user inspection was needed. The overall processing time for 12 day-long recordings varied from few minutes (5-10) to an hour, depending on the number of artefacts. Conclusions: Our seizure detection tool provides high sensitivity, with acceptable specificity, for long and short-term EEG recordings from chronic rat epilepsy models. This has the potential to improve the efficiency and rigor of pre-clinical research and therapy development using these models. Funding: Funding to assist this research program was provided by Kaoskey Pty Ltd, Sydney, Australia