Examination of short binary sequences for mfERG recording

Abstract

The mfERG, when first introduced by Erich Sutter used long sequences with short periods (approximately 15 ms) between steps (flashes). Since then a number of studies have introduced slower or modified sequences to emphasise Oscillatory Potentials (OPs), Optic Nerve Head Components (ONHC) or the s-wave. With this reduction in the rate of presentation many of the investigators have reduced the length of the sequence to allow a shorter recording period. This is also desirable for patient comfort and co-operation in diagnostic investigations. When reducing the length of the sequence there is a risk that it will be too short to ensure orthogonality of the first order component and all significant higher order components, particularly when a large number of areas (hexagons) are stimulated. This paper aims to verify that a short sequence (using the sequence used by the Roland Retiscan stimulating 19 hexagons) is capable of keeping responses of both first and higher orders separate for each stimulating area. The sequence was investigated by placing photodiodes connected to a Diagnosys Espion and then exported to Excel and MATLAB for analysis. It was determined that the sequences used were m-sequences length n = 9. The photodiode only responded to flashes of light so was unable to detect a correcting 0 at the end of sequence. The sequences driving each hexagon were then determined and found to be shifted 26 steps from each other. The correlation coefficients between all sequences was found to be -1/(2n -1). The sequences to decode the second order kernels were determined and the correlation coefficients between each of these sequences, and between these and the original sequences, were also -1/(2n -1). This work provides a mathematical validation of the use of short sequences for slow mfERG, and describes an empirical test method.