Noise-Robust Detection of Hyperbaric Oxygen Seizure Onset

Abstract

This study examined hyperbaric oxygen seizures (seizures induced by breathing pure oxygen under pressure). In addition to the subject's importance to military diving, this preparation is of more generic interest because it provides an experimental model of chemically induced seizures that are generalized at onset. The object of this preliminary study was to find noise-robust numerical procedures that can identify the time of seizure onset. Several candidate methods were compared. They included a high order FIR filter, wavelet denoising and computation of the Hurst exponent. In these calculations, the original signal was corrupted with progressively larger amplitude additive noise. All three methods successfully identified seizure onset to an SNR of -10 dB. Of the methods considered, only the Hurst exponent was able to find the seizure when the SNR dropped to -20 dB.