Detection of neurological injury using time-frequency analysis of the somatosensory evoked potential

Somatosensory evoked potentials (SEPs) can be monitored during critical surgery to help detect or possibly prevent post-operative injury to the brain. This paper presents the application of time-frequency analysis to detect both temporal and spectral changes in the SEP waveform that may occur due to injury. Time-frequency distributions, which provide a measure of signal energy at both a specific time and frequency, were computed for averaged SEPs acquired from anesthetized cats during various stages of hypoxic injury and then recovery. Wigner distributions of SEP waveforms were found to contain a peak of signal energy at a specific time and frequency, a peak that is altered during injury. Four characteristics of the distribution peak that demonstrate changes due to injury were computed: peak time, peak frequency, peak power, and peak sharpness. Peak time was found to increase while peak frequency, peak power, and peak sharpness were found to decrease during injury. Furthermore, the total signal power in a time-frequency space around the normal peak location was monitored by developing a time-frequency window filter (TFWF) method. For all cases, onset of hypoxia was detected an average of 2.75 min earlier by the TFWF method than by the conventional amplitude measurement method. Time-frequency analysis of EP signals may therefore be useful as a monitoring tool for early detection of brain injury.