Abstracts of the XLV International Symposium of ISCEV—Hyderabad, India, 25–29 August 2007

Abstract

Introduction: Most commonly, the estimation of transient evoked potentials such as the ERG, PERG and VEP is by simple ensemble averaging, synchronized by the stimulation signal. This requires three assumptions: (i) the signal to be recovered is stationary; (ii) the noise (or ‘incoherent’ signal component) is stationary with zero mean; (iii) the number of epochs sampled is sufficient to statistically represent the signal of interest within acceptable confidence limits. Whilst ensemble averaging still forms the cornerstone of transient evoked potential measurement, it does not exploit modern statistical signal processing techniques and the potential for extracting more statistical information from the recorded signals. Purpose: To describe a statistical bootstrap method that provides an estimate of the probability that the response obtained is due to random variation in the data rather than a physiological response (viz., the null hypothesis). This method can be applied to (almost) any signal parameter (e.g., power, amplitude range, estimates of signal-to-noise ratio) and is based on randomly re-sampling (with replacement) of the continuously recorded data. Method: The proposed method was developed and tested initially on simulated data with realistic autoregressive moving average (ARMA) noise. This was extended to a series of clinical PERG recordings. Results: The bootstrap model was able to detect the presence of PERG responses at user-defined significance levels in both artificial and clinical recordings. Conclusion: The bootstrap re-sampling technique is simple to implement, very flexible in terms of the signal features that can be statistically analysed, and provides a novel means of objectively testing signal recovery. It thus holds the potential to optimize the acquisition process, and to determine the shortest recording time required, based on a clearly defined statistical criterion.