A fast automated method for calculating the EOG Arden ratio

• Published in: Documenta ophthalmologica Vol. 128; no. 3; pp. 169 - 178
• Main Authors: Sarossy, Marc G., Lee, Matthew H. A., Bach, Michael
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.06.2014; Springer Nature B.V
• Subjects: Electrooculography - methods; Humans; Light; Medicine; Medicine & Public Health; Models, Statistical; Ophthalmology; Original Research Article; Retinal Pigment Epithelium - physiology; Retrospective Studies; Electrooculography; Vitelliform macular dystrophy; Statistical models
• ISSN: 0012-4486; 1573-2622; 1573-2622
• DOI: 10.1007/s10633-014-9430-5

Background Recording of the dark trough/light peak of the electrooculogram (EOG) remains a useful electrodiagnostic tool. Manual analysis of the recording is tedious and lengthy, and automated analysis needs to deal with artefacts due to suboptimal patient cooperation. Methods We present a novel method of automating the processing and analysis of raw EOG data using the open-source statistical software R. Rather than attempting saccade detection, we utilize the fact that basic properties of the response (rough waveform timing) are known and simply fit a square wave to each response run—free parameters are amplitude and phase. To assess this analysis method, responses from 54 eyes of 27 patients with a variety of ophthalmic diagnoses were analysed with manual calculation and with a number of automated methods of fitting the response curve. The Arden ratio was the main outcome measure. Results Robust regression of a fundamental with a three-harmonic approximation of a square wave was found to be the best method. Classification accuracy with this method compared with the manual calculations as gold standard; using a lower normal threshold of 200 %, Arden ratio was found to achieve a sensitivity of 96 % and specificity of 81 %. Time taken to process and analyse the data for a subject was reduced from 20 min for the manual method to 2 min for the automated method. Conclusions The simple approach yielded a surprisingly effective automatic estimation of the Arden ratio. In one author’s laboratory (MB), this procedure has proved to be useful over 5 years for routine analysis.