Abstract
Purpose: :
Continuous Wavelet Transformation (CWT) is an established technique for analyzing electrophysiological signals like electroencephalograms or electrocardiograms, but rarely used for analysis of electroretinography (ERG). One reason is, that in contrast to Fourier Analysis (FA), CWT requires more parameters to be set, rendering the interpretation of results more difficult. To face this problem we are introducing a simple score and provide normative values for the classification of flicker ERG with CWT.
Methods: :
Based on 19 recordings of 16 healthy subjects (age range: 20 to 67 years, mean: 37), normative data for a 30 Hz flicker ERG (4 ms flash, post-trigger time 200 ms, sampling rate 2 kHz, integrated stimulus luminance 2.25 cds/m² white, background luminance 30 cd/m²) have been established. Sweeps were detrended and averaged using the Tübingen ERG Explorer and CWT was applied using SeaSolve AutoSignal to calculate the matrices of coefficients in time-frequency domain. 95%-quantiles for each element in the matrices were calculated. In the resulting matrix 12 spots representing peaks and troughs of the original signal were identified. Using these spots the implicit times (defined by the center) and the surface integral (defined by a circle of a fixed radius), were determined and used as score for classifying recordings. Clinical applicability was evaluated comparing recordings from patient groups (CSNB, Retinoschisis, DR, CAR, MAR). Results were assessed for significance in comparision to noise by using 1st-order auto-regressive red noise estimation.
Results: :
Values of the surface integrals showed a very good correlation to the values obtained by FA for the 30 Hz bin (r=0.99). In patient groups the score showed an expected attenuation. In comparison to FA, CWT was able to provide additional information and showed clearly distinct patterns of frequencies during peaks and troughs of the flicker recording. During peaks usually more and higher frequencies are involved compared to troughs. Diseases result in modified waveforms like CSNB (double-peaks), DR and artery obstructions (flattened peaks/troughs) showed reduced patterns compared to the normative data.
Conclusions: :
The results indicate that analysis using CWT rather than FA has the advantage of providing additional information about changes of the frequency distribution in time. This information may be used to further classify ERG recordings, using a scoring system as described here. Combined with significance estimation and normative data this may allow for a more accurate interpretation than merely summing amplitudes or averaging peak-to-trough. CWT has the ability to unveil yet unknown coherences.
Keywords: electroretinography: non-clinical • electrophysiology: non-clinical