Purchase this article with an account.
A. A. Foulis, S. Parks, D. Keating; A Multilayered Approach to the Automatic Analysis of MFERG Waveforms: Analysis of the Fourier Domain Profile. Invest. Ophthalmol. Vis. Sci. 2009;50(13):4517.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The multifocal ERG (mfERG) provides spatial and temporal information on the retina’s function in an objective manner making it a valuable tool for diagnosing abnormalities of the retina. Interpretation of these signals can however be difficult and subjective, sometimes resulting in inconsistencies in the interpretation process. Consequently, an objective means of determining the presence of a response, and where appropriate, automatically analysing the waveform would be advantageous. A number of approaches are being investigated to provide an expert system designed to ease and improve the interpretation process and to provide a more consistent method of analysis. These include the use of artificial neural networks, signal to noise ratio mapping, signal processing techniques and analysis of the signals in the Fourier domain.
This study describes the assessment of Fourier profiles, both of the uncorrelated data and of the correlated data. Firstly the uncorrelated data signal from recordings representing a wide range of conditions and patient compliance was examined in the frequency domain. This was compared for each patient to establish if the quality of a recording and the presence of retinal function could be determined from the Fourier profile. For more localised information, the correlated data was transformed into the Fourier domain. The frequency distribution was analysed to establish if a difference between true responses and those with no function could be detected.
The frequency profile of the uncorrelated data signal showed distinctive components: a normative type frequency distribution was observed in the presence of retinal function while low frequency noise, which could be attributed to artefacts such as blinking and eye movement was also seen. The prevalence of this low frequency noise was dependent on the quality of the recording: it increased for less compliant patients. The frequency spectrum of the correlated data showed that a key indicator of a good individual signal was a distinctive distribution centred at half the driving frequency (although this is specific to the digital style LCD stimulus used here).
The Fourier profile of the uncorrelated data signal provides information regarding the overall integrity of the recording and differentiates the noise from the signal. Assessment of the correlated signals is an objective method of determining if a specific signal should then go onto further stages of the expert system.
This PDF is available to Subscribers Only