Objective detection of visual field defects with multifrequency VEPs

Abstract

Purpose: To correlate multifrequency pattern reversal VEPs in quadrants (QmfrVEPs) with perimetric field losses for objective detection of visual field losses.

Methods: QmfrVEP measurements were performed using four LED-based checkerboard stimulators to stimulate the four quadrants of the visual field. QmfrVEPs were measured monocularly in 5 normal subjects and in 5 glaucoma patients who showed losses in conventional Octopus perimetry. The pattern reversal frequency varied slightly between the stimulators: (11.92, 12.00, 12.08 and 12.16 reversals/sec). The responses to the different stimuli were identified by discrete Fourier analysis. VEPs were recorded using different electrode configurations, and the recording with the highest signal-to-noise ratio (SNR) was used for further analysis.

Results: QmfrVEP responses from the different quadrants can be reliably measured and separated using the 0.08 reversals/sec interstimulus reversal frequency differences. The signal-to-noise ratio in the four quadrants was significantly correlated with the equivalent visual field losses obtained with perimetry (Spearman rank correlation: P < 0.001). In the five glaucoma patients, the SNR was reduced in 15 out of the 16 quadrants with a perimetric defect, in comparison to the results in quadrants of healthy subjects. This confirms the sensitivity of the procedure.

Conclusion: QmfrVEP responses can be measured reliably. This pilot study suggests that high SNR values exclude visual field defects and that focal defects can be identified in glaucoma patients.

Trial registration: www.

Clinicaltrials: gov . NCT00494923.

Keywords: Objective visual field test; Pattern reversal; Quadrant multifrequency VEP; Signal-to-noise ratio; Steady state.

Fig. 1

Experimental setup. The subject's monocularly focuses on a dim red LED, which is located in the middle of four arrays

Fig. 2

A Field of stimulation at a viewing distance of 32 cm. The side length of the arrays is 20.8°. The frame adds up to 1.9°. The arrays are denoted: Upper Left = UL, Upper Right = UR, Lower Left = LL, Lower Right = LR. B Amplitude spectrum of one channel (CH: 7) for a measurement with quadrant stimulation in a normal eye. The stimulus frequencies in the four arrays were: UL = 12.00 Hz, UR = 11.92 Hz, LL = 12.08 Hz, LR = 12.16 Hz. The amplitudes UR, UL, LL and LR correspond to the responses from the stimulus frequencies at the four arrays. The smaller amplitudes next to the stimulus amplitudes represent the random noise that was used for the SNR calculation

Fig. 3

The present setup used five head electrodes and 7 channels. The electrodes A–C are located 4 cm right, left and above the Inion. Electrode D is fixed 2 cm beneath the Inion. An additional electrode E is placed 4 cm above electrode A. Dashed lines show virtual channels (5–7), calculated by real measured channels (1–3). The percentage of the records in which the concerning channel displayed the largest SNR is also shown

Fig. 4

VEP and perimetry in a glaucoma patient. The figure shows screen prints of the present VEP software and Octopus G1. A Octopus perimetry of the central retinal area (12°) of the patient. Perimetry reveals advanced damages in the upper left quadrant. Here, defect depths of up to 30 dB are present. Also, in the upper right and lower right quadrants defects are observable, while the lower hemifield shows only minor perimetric defects. B "Results summary" of the present programme presents results for all LED arrays. For all quadrants, the "best-SNR" as well as the corresponding amplitude and phase is given. The channel from which the "best-SNR" has been derived is depicted in brackets. The columns at the side of each field indicate the level of the SNR graphically. In correspondence with decreasing perimetric losses (Fig. 4A), the SNR increases clockwise starting in the upper left quadrant (UL) until being the largest in the lower left quadrant (LL)

Fig. 5

Spectra for 7 channels show high amplitudes for lower stimulus areas with normal visual fields. Amplitudes corresponding to the upper quadrants revealing perimetric defects are smaller

Fig. 6

The scatterplots show the SNR results of the four quadrants individually for the five normal subjects (top row) and the five glaucoma patients (bottom row) as functions of the visual field defects. The vertical lines indicate the cutoff that separates “damaged” and “not damaged” quadrants. The other lines show the predictions from the linear mixed-effects model that was fit to the data. This model accounts for a fixed relationship between perimetry and VEP (the slope is equal for all patients), but also for a random overall variability in SNR between patients (variable intercept = patient effect). Note the considerable individual variability in the overall SNR. A clear relationship between perimetry and VEP can be observed in four patients, and the p-values were < 0.001 for the fixed effects