An Objective Evaluation Method of Multifocal Electroretinogram Amplitude Based on Signal Detection Theory

Abstract

Purpose: : To analyze the distribution of signal and noise magnitude of the multifocal electroretinogram (mfERG) first order kernel. To develop an objective method to discriminate normal amplitude mfERG from low amplitude mfERG.

Methods: : mfERGs were recorded from 10 healthy subjects (25.4 ± 3.5 years-old) and the 103 first order kernel was extracted using the Veris Science 6.05. The magnitudes of a signal time interval and a noise time interval were estimated by calculating the RMS amplitude ratio and dot product from these time intervals with signal time interval from a template. Signal and noise magnitude distributions varied following Gaussian functions. Limit values were estimated for signal and noise magnitudes that covered 99% of signal and noise magnitude distributions. We artificially attenuated the amplitude in the signal time interval of 15%, 50%, and 85% of the 103 waveforms at five different levels. The attenuation consisted of progressively inserting increasing amounts of noise in the signal time interval. The attenuated waveforms were then evaluated by a signal-to-noise protocol that took into account the lower limit of confidence of the mean signal limit of the sample. We quantified the true positive rate and false positive rate returned by the evaluation protocol.

Results: : RMS amplitude ratio metrics. A true positive ratio above 97% was found for all levels of waveform amplitude attenuation tested. The false positive ratio remained between 1-2% for the attenuation levels 2-5, but in attenuation level 1, it raised to 18-20% across the different percentages of attenuated waveforms. Dot product metrics. The true positive ratio remained above 97% for the different percentage of attenuated waveforms at the lowest attenuation level and raised to 85-95% for the next four attenuation levels. The false positive ratio increased with the degree of attenuation for all percentage of attenuated waveforms, ranging between 9 and 50% for the attenuation levels 2-5.

Conclusions: : The evaluation method based on the RMS amplitude ratio was more efficient to discriminate attenuated waveforms than the method based in the dot product. This method can be useful for clinicians to perform an objective evaluation of the mfERG first order kernels.

Keywords: electroretinography: clinical • retina