We propose a study of time-varying evoked potential (EP) signals. Traditionally signal averaging has been used to extract EP signals from random background noise. Averaging may require hundreds or thousands of ensembles making it impossible to follow time-varying changes in EP's. We propose to employ an adaptive filtering algorithm to improve the aignal-to-noise ratio (SNR) for rapid EP processing. The adaptive filter works by recursive minimization of mean-squared error between successive ensembles. In preliminary studies with simulated data we show that adaptive filtering significantly improves the SNR. Preliminary work on processing human EP data shows that adaptive filtering requires significantly fewer ensembles than averaging to attain comparable SNR. We wish to test the following hypotheses: (i) Adaptive filtering is a method superior to ensemble averaging for studying transient changes in EP amplitude and latency. (ii) We will use hypoxic animal models to test the hypothesis that rapid EP processing may serve as a diagnostic test of cerebral ischemia. (iii) Continuous monitoring of transient EP phenomena would be useful in some specific cases in neurologic intensive care and surgery. We propose to (i) further develop the theory of adaptive filteirng, develop new algorithms, and design a special purpose digital signal processing computer. (ii) We will employ the signal processor, in conjunction with stimulators and low-noise amplifiers for rapid and continuous monitoring of somatosensory evoked potentials (SEPs) and brainstem auditory evoked response (BAER) from animals and humans. In animal studies we will monitor transient changes in SEPs that occur in conjunction with induced hypoxia and cerebral ischemia. We will assess the utility of monitoring and rapid processing of EP in neurosurgery and neurologic intensive care. We hope to show that monitoring transient EP phenomena may serve as a diagnostic tool for trauma to brain due to oxygen deprivation, head injury, disease or surgery.
Showing the most recent 10 out of 53 publications
