Sampling and interpolation of the a-wave of the electroretinogram.

UNLABELLED: This study was undertaken (a) to determine the minimum sampling frequency required to record a-waves evoked by flashes of very high energy without significant distortion and (b) to demonstrate that data sampled at a minimally adequate frequency can be interpolated to reconstruct the original waveform.
METHODS: Dark-adapted ERGs from two anaesthetized macaque monkeys and an adult human were studied. Responses evoked by high-energy flashes that produced a-waves peaking as early as 5 or 6 ms after the flash were sampled at 5 kHz and transformed to obtain their discrete Fourier spectra. The amplitude of all spectral components above some cut-off frequency (e.g., 400 Hz) was then set to zero and the modified spectra transformed back into the time domain. The resulting computed responses, which contained no Fourier components above the cut-off frequency, were compared with the original recorded samples. To assess the validity of one method of interpolation, sample sets consisting of every fifth point of records sampled at 5 kHz (i.e., sets of 1 kHz samples) were subjected to Fourier transformation to give spectra with a frequency range of 0-500 Hz. These spectra were extended from 500 Hz up to a much higher frequency (e.g., 8 kHz) by adding zeros. The extended spectra were transformed back into the time domain to provide sets of interpolated samples at twice the chosen spectral frequency limit (i.e., 16 kHz).
RESULTS: Removing all Fourier components above 400 or 500 Hz had no significant effect upon the leading edge or peak of the a-wave. However, removing Fourier components above 500 Hz gave rise to slight distortion of the oscillatory potentials (OPs) that appeared just after the a-wave peak on the leading edge of the b-wave. Except for this small distortion, the original 5 kHz data samples corresponded very well with the interpolated curves that had been generated as described above from a 1 kHz subset of the samples. This provides further confirmation that dark-adapted ERG a-waves evoked by flashes of up to about 50,000 sc. Td sc do not contain Fourier components with frequencies above 500 Hz.
CONCLUSION: Human and macaque a-waves are completely represented by 1 kHz samples and Fourier methods can be used to reconstruct the original continuous waveform. However, to capture the OPs with complete fidelity, a higher sampling rate is necessary.