The effects of bandpass filtering on the oscillatory potentials of the electroretinogram

In order to study the OPs, the ERG signal must be filtered to eliminate the low-frequency waves known as the a- and b-waves. Unfortunately, the ISCEV ERG standard does not give clear guidelines on how to proceed apart from indicating that frequencies below 75 Hz should be filtered out when recording scotopic OPs, while no suggestions are offered for the photopic OPs. The purpose of this study was thus to characterize more extensively the effects of various digital filters on the photopic OP waveforms in order to suggest the most appropriate filtering method to record them.

Filtered OPs (N = 9600 tracings) were extracted from a photopic ERG databank of 40 normal subjects [intensity: 4.4 cd s m^-2; background: 30 cd m^-2] using 240 different combinations of five digital filters types (Bessel; Butterworth; Elliptic; Chebyshev type 1 and 2), eight bandwidth ranges (50-300; 75-300; 100-300; 125-300; 50-1000; 75-1000; 100-1000; 125-1000 Hz), three filter orders (1, 2 and 5) and with/without phase lag corrections that were generated using MATLAB 2015b. The peak time and the percentage of OPs (sum of OP amplitudes on the b-wave amplitude) were calculated in the time domain (TD%OP).

The timing of the OPs was less affected than the amplitude by the different filters used. Depending on the filter used, the resulting OPs were either severely depressed (16.16% of broadband OP content) or slightly reduced (93.63%). The filters that most successfully eliminated the slow components of the ERG (i.e., < 12% of broadband value) were the Bessel, the Butterworth and the Chebyshev type 1 filters and out of the latter, the Butterworth filter was that which most faithfully reproduced the high-frequency OPs (i.e., > 96%).

Our results vividly demonstrate the need to better define the characteristics of the filter that is used to record the OPs as it does have a significant impact on the resulting waveform.