Signal to Noise ratio in multifocal ERG records.

Abstract

Abstract: : Purpose:There are a number of factors which influence the signal to noise ratio (SNR) of multifocal responses. A new method of measuring the SNR was used to evaluate signal quality. Methods: In multifocal systems, an appropriate m–sequence is usually decimated into 64 or 128 columns to give a set of orthogonal sequences to control the individual elements of the stimulus. If decimation is by 64 and the stimulus has 61 elements then this leaves three sequences, ‘dead sequences’, which are not correlated with any element. These ‘dead sequences’ can therefore be used to obtain a direct measure of the noise. Six subjects performed m–sequence controlled global flash experiments using an LED stimulator. The global flash gives significant improvement in SNR enabling sub–components of the response to be recovered. The sequence length was varied from 12 through to 15 and the driving frequency varied from 10 to 500 Hz. Results:The SNRs obtained from the Global LED stimulus ranged from 21 to 91 as the m–sequence degree was varied from 12 to 15. For a 12 bit sequence, the SNR at 500Hz is only 10% of that obtained at 10 Hz. Conclusions:The technique of cross–correlating a ‘dead sequence’ with the raw data signal produces robust measures of SNR without having to look at different time windows in the cross–correlated response. Sequence length, stimulus delivery method and stimulus driving frequency have significant effect on the SNR of the multifocal ERG. Simple signal averaging theory predicts a √2 improvement in SNR as the m–sequence degree is increased by one and these results show good agreement with the theory.

Keywords: electroretinography: non–clinical • electroretinography: clinical