Abstract
Purpose: :
Retinal steady-state potentials play an important role, especially when evaluating very low-voltage retinal responses (sub-microvolt ERG). Extraction of Fourier components enhance reliability, through intrinsic immunity to noise. Recent studies address regulation processes in the retina, using sustained pattern or flicker stimulation to detect underlying trends in the neural response [1,2], thus further challenging signal statistical significance. To this purpose a new test was developed, and its usefulness checked in a realistic environment.[1] Porciatti et al. Invest Ophthalmol Vis Sci. 2005 Apr;46(4):1296-302.[2] Fadda et al.Clin Neurophysiol. 2009 Oct;120(10):1828-34.
Methods: :
Sustained PERG or FERG responses at 8Hz where recorded in 5 normal subjects, using a verified acquisition setup. Recordings were divided into 20 segments of 60 cycles. Fourier components of signal and noise were computed for every segment and made available for subsequent off-line analysis. Theoretical work was carried on with MathCAD and MatLab software packages.
Results: :
A probability density function (pdf) of the average of N independent measurements of noise amplitudes was found, assuming a Rayleigh pdf of base noise. This parametric function shows very good accuracy when N>4, converging to an exact limit function as N increase. Distribution of S/N ratio was then computed, and tables of confidence levels evaluated for a set of p and N values (ex: N=20, p=0.05, S/N>2.03). An assessment of experimental data was then made, using the proposed model. It was found that a p=0.05 level was attained in most experiments, while alternative statistical approaches failed to demonstrate similar significance.
Conclusions: :
The proposed method is statistically powerful, and results approach the theoretical limits. Statistical significance was verified in an experimental study of 7.5 s segments over a long temporal series, but the method is useful in more general conditions, being able to determine the temporal resolution that may be achieved in an evaluation of ERG adaptive changes.
Keywords: electroretinography: clinical • electrophysiology: clinical • computational modeling