Seizure detection on/off system using rats´ ECoG

We present an enhanced algorithm for seizure onset and offset detection in rats´ ECoG. Because a seizure in rats´ ECoG evolves much more stereotypically than that in human, analyzing seizure evolution in rats´ ECoG is advantageous to understanding the evolution process. The proposed algorithm outperforms a prior automatic seizure detection and termination system in in-vivo rats´ ECoG. We improve the algorithm by using relevant frequency bands of 14-22 Hz to onsets and 7-45Hz to offsets; by using spectral power rather than spectral amplitudes for its feature; and by replacing the 2-point moving-average filter for postprocessing with a 2^nd order Kalman filter. Not only does the proposed algorithm provide better detection statistics, but it lowers the system´s complexity by no longer requiring computation of a fast Fourier transform and by using a single structure with the two different spectral power features for onset and offset detection.