Premium
Recovery of the ERG from very low SNR recordings using a robust frequency domain approach: an Internet open source implementation
Author(s) -
FISHER A,
ELEUTERI A,
TEYMOURI R,
HAGAN RP
Publication year - 2012
Publication title -
acta ophthalmologica
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.534
H-Index - 87
eISSN - 1755-3768
pISSN - 1755-375X
DOI - 10.1111/j.1755-3768.2012.2424.x
Subject(s) - erg , noise (video) , computer science , estimator , algorithm , statistics , speech recognition , mathematics , artificial intelligence , physics , optics , image (mathematics) , retina
Purpose To present a statistically robust method of estimating the transient electroretinogram (ERG) in extremely poor recordings (very low SNR) contaminated by continuous noise (electrode noise, EMG, exogenous electrical interference) and discontinuous noise (eye‐movements and blinks). The transient ERG (ERG) is estimated by ensemble averaging with standard errors and p value, derived by either Magnitude Square Coherence or Bootstrap Resampling. However, both techniques are susceptible to noise. A robust estimator noise is required. Methods A series of ERG’s was recorded in normal subjects, containing numbers of prescribed eye movements and blinks. AR models were derived using the Burg algorithm over a range of explicit SNR’s. The ERG record was recovered as its Fourier Series representation over 9 harmonics. p values were derived for all estimates. An Excel implementation is presented which accesses remote Internet resources using MatSOAP©. Results The robust estimator was validated in Monte Carlo realisations of the AR model with expected high levels of performance for amplitudes and p values. In estimations in the raw data sets, high levels of statistical agreement were achieved for each subject across no‐noise:noise series. The recovery by ensemble averaging performed poorly. Conclusion The robust estimation in the ERG as presented here is novel and offers an important advance in clinical ERG measurement. The complexity is hidden from the User by providing a familiar Excel© spreadsheet interface.