Documenta Ophthalmologica 97: 157–178, 1999. © 1999 Kluwer Academic Publishers. Printed in the Netherlands.
A comparison of two multifocal ERG systems M. BOCK, M. ANDRASSI, L. BELITSKY & B. LORENZ Department of Paediatric Ophthalmology, Strabismology and Ophthalmogenetics, University of Regensburg, Germany Accepted 23 June 1999
Abstract. The First Order Kernel responses (FOK) from a group of 15 normal subjects aged 21–36 yrs were recorded under clinical conditions with the multifocal ERG (MERG) systems VERIS and RETiscan using system-specific software and hardware under various parameter settings. First, the multifocal ERG’s of normal subjects were recorded with the standard-parameters of VERIS and RETIscan. Thereafter the VERIS system parameters were set as close as possible to the RETIscan-setup, and the recordings were repeated. The VERIS parameters changed were the luminance- and contrast-settings of the stimulus-monitor and m-sequence-timing. In addition, the effects of different sampling rates were also examined. The consequences of the parameter changes were analyzed by displaying the peak times of the First-Order-Kernel (P1). The parameter-combinations used for recording lead to systemspecific results. By varying the luminance/contrast settings and/or the m-sequence-timing the results can be changed. The amplitude of the recorded results can be influenced by the luminance-settings of the stimulus-monitor, and the topography of the P1 peak times is affected by the m-/f-frame-combination of the recording stimulus. With their standard parameters, the two systems give different measures of the P1 peak times. By adjusting the recording parameters, these differences can be reduced significantly. The main parameter here is the msequence-timing, although the sampling rate must also be taken into account with regard to the accuracy of the recording results. Key words: electroretinogram, multifocal ERG Abbreviations: MERG – Multifocal ERG; FOK – First-Order-Kernel; N – Number of subjects recorded per experiment
Introduction The multifocal-technique introduced by Sutter and Tran in 1992 allows recordings of multiple spatially resolved ERG-responses from the retina over a central area of about 25 degrees using an m-sequence-based flicker-stimulus that consists of many inputs [1, 2]. The ERG data is recorded with one channel and the responses are derived from this summed signal with the help of the cross-correlation-technique. The results can be found in different kernels, for example a linear approximation of the system in the First-Order-Kernel (FOK) [1, 2]. The recording results are analyzed in different ways, e.g. using
158 concentric-ring-grouping with evaluation of peak times and amplitudes, combinations of both or by displaying topographies of latencies and amplitudes [1, 3–6]. These analyses are important in describing the phenotype in retinal pathologies. For example, significant changes in both amplitudes and implicit times with eccentricity have been shown in RP [5, 7]. In Stargardt’s disease both amplitudes and implicit times are significantly changed in the center only [8]. The multifocal-systems VERIS and RETIscan differ from each other in several ways. The differences are both hardware and software related. The aim of this study is to compare the performance of VERIS and RETIscan under clinical conditions. Thus both systems have been used with the software and hardware which is provided by each manufacturer. This means that different monitors were used to generate the stimulus, different electrodes, amplifiers and recording parameters, which limit the ability to make comparisons under equal conditions between both systems. However, the variation of single parameters of the recording protocol (experiments B to E) have been realized with one single system (VERIS). Thus the results of single experiments can be estimated individually and finally be related to the RETIscan outcomes up to a certain degree. It is also important to note that the differences due to the software options may have changed if a newer version has been released. In most cases, the changes of the software refer to additional options for parameter-settings and/or data analysis. An example for such a change is the sampling rate which was 480 Hz with our test-RETIscan-system and which has been changed to a variable control parameter in recent software versions. In our experiments which were carried out in November and December ’97 (RETIscan) and from January to March ’98 (VERIS) we used the VERIS-Software 3.01 and the RETIscan-Software 3.0g. Differences of VERIS and RETIscan In order to describe the differences of the systems one can divide the multifocal-technique into several aspects. One aspect is the stimulus-monitor-combination that produces the flicker stimulus. In this context, the m-sequence length and the m/f-frame combination used is a very important factor. Another aspect can be found in the electrode type used, e.g. DTL, Jet or BurianAllen. A third is the amplifier and AD converter section where the continuous signals are converted to sampled curves. The data analysis using the crosscorrelation-technique is used in both systems whereas the artifact handlings differ from each other as well as the stimulus timings [1, Roland Consult, personal communication].
159 Stimulus monitor VERIS uses a Nortech monitor (‘URH21L’) to display the flicker stimulus. The white-level used in this study is 135 cd/m2 and the black-level is 0.1 cd/m2 . This is not the maximum that the monitor is capable to emit, but it produces a stimulus that both normals and patients are tolerating, and with which good signal-to-noise-ratios are observed. The framerate of the flicker stimulus is 75 Hz. The monitor of RETIscan used in this study was an ‘Elsa Ecomo 20S96’ monitor, the luminance- and contrast-controls were both set to 100% resulting in a luminance of the white hexagons of 106 cd/m2 and of the black ones of 4.5 cd/m2 . In combination with the computer that generated the stimulus picture a framerate of 60 Hz was used. All luminance measurements were taken with a ‘Gossen Mastersix’ instrument. Electrode and amplifier With the VERIS system Burian-Allen electrodes were used because of the better signal-to-noise-ratio compared to DTL electrodes. The Burian-Allen electrodes could not be used with RETIscan because the electrodes did not work with the RETIscan amplifier, which is DC based. The RETIscan system works with JET electrodes as well as with DTL electrodes. Comparative measurements were performed with both DTL and Burian-Allen electrodes on the VERIS system, thus DTL electrodes were used in the present study with the RETIscan system. In combination with the shielded cables that RETIscan uses there were no problems with the line frequency of 50 Hz. As there were some noise problems with the line frequency in our work with VERIS, we generally use the additional line frequency filter of the Grass amplifier in our First-Order-Kernel VERIS setup. AD-converter To be able to analyze the data, the analog signals of the recordings have to be converted to a digital signal. The important parameter of the sampling sequence is the time period between two samples. This parameter is given with the sampling rate. VERIS has an option to change the samples per frame and thus the sampling rate. Our standard is 16/frame (1200 Hz) or 0.83 ms between two samples. The RETIscan system used a constant sampling rate of 8 samples/ frame (480 Hz) or 2.1 ms between two samples.
160
Figure 1. The m-sequence-timing of the standard VERIS- and RETIscan-setups, m/f-frame-combinations.
Artifact handling VERIS reduces the artifact level of the recordings with the help of transformation and calculation of time compressed noise signals and time distributed ERG signals into distributed noise information and compressed ERG data in the kernel environment [1]. RETIscan uses online monitoring which means that while measuring the raw data signal is observed. Each time during the measurement when a certain level of amplitude is exceeded, the momentary m-sequence-steps are repeated until the responses have been recorded properly (RolandConsult, personal communication).
M-sequence-timing In the VERIS standard protocol used in this study, there are no f-frames (fill frames) in the m-sequence which means that the active m-sequence-steps follow each other directly every 13.3 ms (Figure 1). The standard VERIS m-sequence exponent is 14 or 15, resulting in 16383 or 32767 m-steps. RETIscan used a special setup for the First-Order-Kernel analysis, which consists of active m-sequence-steps (m-frames, displayed white in Figure 1, msequence exponent=9) and f-frames (displayed black in Figure 1). So every active m-sequence-step lasts 16.6 ms (1 m-frame), and is followed by an interval rest of 66.7 ms (4 f-frames) before the next active m-sequence-step is initiated. This means that the total time from one active m-sequence-step to the next is at least 83.3 ms (RolandConsult, personal communication).
161 Methods The First-Order-Kernel responses from a group of 15 healthy subjects aged 21 – 36 yrs (visual acuity between 1.0 and 1.4) were measured one time with RETIscan and several times with VERIS using various parameter settings. This included one run with the standard parameters of each system. Additionally, three runs with mixed parameters were measured in subsets of four to thirteen subjects with the VERIS system. In those experiments, only one single parameter of the recording protocols was changed in order to investigate the consequences to the recording results. Informed consent was obtained from all subjects after explaining the recording procedures. Tenets of the declaration of Helsinki were followed. The refractive errors of the subjects were corrected for testing distance. The parameter settings were varied in order to adapt the VERIS system to the standard parameters of RETIscan or to analyze how the result of the MERG is changed when one parameter of the setup is altered. To obtain a good signalto-noise-ratio Burian-Allen electrodes were used with the VERIS recordings, with RETIscan DTL electrodes (monopolar electrode in form of a thread, additional reference point needed) were used because the RETIscan amplifier did not allow the use of Burian-Allen electrodes. Previous experiments with Burian-Allen- and DTL electrodes on the VERIS system had shown that the performance of the electrode types is similar except for the signal-to-noiseratio. This is illustrated in Figure 2 where the P1 amplitudes and P1 peak times of a control group of 20 subjects (aged 16 – 52, visual acuity 1.0 – 1.4) recorded with both Burian-Allen and DTL electrodes are displayed (VERISsetup: 103 hexagons, recording time 4 min (Burian-Allen electrode), 8 min (DTL electrode), distance eye-monitor = 30 cm, subjects dilated and corrected for refractive error in testing distance). These subjects were not identical to the subjects used in the VERIS/RETIscan comparison study below. The amplitudes and the peak times of the concentric rings 2, 4, 5 and 6 (refer to Figure 2, central hexagon = ‘ring’ 1, surrounding hexagons = rings 2 – 6) are almost similar. The reason for the minor differences in the P1 peak times of ring one and ring three is the noise in the DTL recordings and the decreased number of hexagons averaged in the concentric ring analysis (ring 1: 1 hexagon, ring 2: 6 hexagons, ring 3: 12 hexagons, ring 4: 18 hexagons, ring 5: 24 hexagons, ring 6: 42 hexagons). The responses to the stimuli are correlated to the timing of the stimuli, but the noise is a random signal uncorrelated to the stimulus timing. Thus by averaging the real response becomes clearer and the noise disappears when a certain amount of single responses are added together and nominated afterwards. So the signal-to-noise-ratio is getting better with the number of hexagons averaged to the result. Therefore
162
Figure 2. The concentric grouping (6 concentric rings of an 103 hexagon pattern) and boxplots of P1 amplitudes and P1 peak times (using concentric grouping) of 20 subjects aged 16-52 yrs measured with Burian-Allen- (on the left) and DTL-electrodes (on the right) on a VERIS-system with standard parameter settings.
163 responses from the inner areas (for example ring one to three) are much more sensitive to noise peaks. Those peaks can lead to different local extrema and therefore to wrong estimation of the maximal amplitudes and peak times. The smaller signal-to-noise-ratio of the DTL electrodes in our RETIscan measurements in this study was balanced by using a hexagon pattern of 61 hexagons (VERIS standard: 103 hexagons), because the signal-to-noise-ratio improves by increasing the signal with a larger stimulus. Both systems used stretched stimulus pictures with larger hexagons in the periphery than in the center. Altogether, five different experimental setups were employed (A to F): A. Standard-results of VERIS and RETIscan The 15 subjects of the control group were measured with RETIscan using the following parameters: stimulus monitor luminance settings: 106 cd/m2 (white), 4.5 cd/m2 (black). The MERG’s were recorded with DTL electrodes, a 61 hexagon pattern and amplifier cut-off settings of 10 and 100 Hz. Each recording was split into 8 runs (m=9 in each run), resulting in a total recording time 5:40 min. The m-sequence-timing of Figure 1 (right side) was used. The framerate was 60 Hz and the sampling rate was 480 Hz. In addition, the 15 subjects were recorded with our VERIS standard recording protocol with Burian-Allen electrodes, 103 hexagons (displayed on top of Figure 2) and a standard m-sequence-length with m=14 (m-sequencetiming of Figure 1, left side) resulting in a total recording time of 3.38 min. The stimulus-monitor-luminance-settings were 135 cd/m2 (white) and 0.1 cd/m2 (black), the amplifier cut-off-frequencies were set to 10 and 300 Hz (with an additional line frequency filter at 50 Hz), the framerate was 75 Hz and the sampling rate was set to 1200 Hz. For both systems all subjects were dilated and corrected for the testing distance of 30 cm. B. VERIS results with an adapted setup In this experiment a VERIS setup matching the original RETIscan protocol as close as possible was employed. The following VERIS parameter settings were used: luminance settings of the stimulus monitor: 106 cd/m2 (white), 4.5 cd/m2 (black), Burian-Allen electrodes, 61 hexagon pattern, amplifier frequency settings 10-100 Hz (additional line frequency filter), m=12, resulting in a total recording time of 5.27 min., m-sequence-timing of Figure 3 (m:f-frame = 1:5), subjects dilated and corrected, distance to screen 30 cm, framerate 75 Hz, sampling rate 1200 Hz. The most important change of the
164
Figure 3. A comparison of the original RETlscan-m-sequence-timing and the adapted VERIS-m-sequence-timing.
VERIS setup was the slower rate of stimulation due to the change in the m-sequence-timing (Figure 3). To create a VERIS-stimulus that matches the timing of the RETIscan stimulus as closely as possible, 5 f-frames (displayed black in Figure 3) were inserted after each active m-sequence-step (displayed white in Figure 3) resulting in a nearly 80 ms interval period between 2 active m-sequencesteps. The number of m-steps was the same as with RETIscan, where the sequence was run 8 times to improve the signal-to-noise-ratio. So while RETIscan was working with a short m-sequence of eight times 512 steps (in total 4096 steps), VERIS used a longer m-sequence of 212 –1=4095 steps. The differences between the systems and the setups that remained were different hardware (for example monitor and amplifier), the flicker framerate (RETIscan: 60 Hz, VERIS: 75 Hz), the electrode used (RETIscan: DTL; VERIS: Burian-Allen), the different sampling rate (RETIscan: 480 Hz, VERIS: 1200 Hz) and the different artifact rejection methods. C. Influence of the different stimulus monitor settings To estimate the effect of different luminance and contrast settings of RETIscan and VERIS caused by different stimulus monitors another experiment was made with 4 subjects aged between 21 and 31 yrs (visual acuity between 1.0 and 1.2) of the control group. Each of the four subjects was measured on VERIS with the same parameters as in (B) except for the contrast and luminance settings of the stimulus monitor. Three different settings of the monitor were used: 1. VERIS standard white-level (100% of the VERIS monitor luminance settings or 135 cd/m2 ), 2. RETIscan standard white-level
165 (85% of the VERIS monitor luminance settings) and 3. another luminance (70% of the VERIS monitor luminance settings). The black level was set to the standard luminance of RETIscan in all combinations (4.5 cd/m2 or 23% of the VERIS luminance control settings). By simulating the contrast and brightness settings of the RETIscan stimulus with VERIS the hardware characteristics and non-linearities of the different stimulus monitors of both systems were neutralized. D. Influence of the m-sequence-timing To show the importance of the m-sequence-timing used in the VERIS setup with respect to the results of the P1 peak times (A) and (B) of the VERIS recordings one subject aged 28 yrs (visual acuity 1.2) was recorded with VERIS two more times and with exactly the same setup parameters except for the m-sequence-timing. For the first recording the standard m-frame-setup of VERIS (Figure 1) was used and for the second recording the adapted VERIS m-sequence-setup (Figure 3) was used. The remaining parameters were kept as in (B). Again using the same hard- and software of the VERIS system ensures that the changes of the recorded results must be related to the setup parameter which settings were altered. Both measurements were recorded within 10 minutes to avoid changes of the recording conditions. E. Sampling rate To estimate the effect of different sampling rates, 6 of the 15 subjects of the control group (aged 21 – 34 y) were measured with VERIS and the parameter setup of experiment (B) with exactly the same parameters except for the sampling rate, which was set to 450 Hz, 1200 Hz (VERIS standard) and 2400 Hz. The MERG’s were recorded consecutively within 20 min to minimize changes of the recording conditions.
Results A. Standard-results of VERIS and RETIscan In Figure 4 typical trace arrays of recordings from the same subject are displayed. In the VERIS trace array an artifact rejection with one number of iteration was used. In the RETIscan trace array the curves have been smoothed once (for the analysis of the P1 peak times showed later on there was no smoothing). Because of the differences of the hardware and the recording protocol (for example, the different number of hexagons) the results can not
166 Figure 4. The trace array and concentric ring grouping of a VERIS recording and a RETIscan recording from the same subject: VERIS recording (on the left): normal subject aged 27 yrs; parameters: 103 hexagons, Burian-Allen electrode, recording time 3.38 min, artifact rejection with one-time-iteration. RETIscan-recording (on the right): same subject; parameters: 61 hexagons, DTL electrode, recording time 5.40min, one time smoothing.
167
Figure 5 (a). The boxplots of the P1 peak times for standard recordings with VERIS main lines of the boxplots at 25th, 50th, 75th, further lines at 5th and 95th percentiles (number of subjects: N=15).
Figure 5 (b). The boxplots of the P1 peak times for standard recordings with RETIscan, main lines of the boxplots at 25th, 50th, 75th, further lines at 5th and 95th percentiles (number of subjects: N=15).
168 be compared. The corresponding concentric ring analysis for each trace array is also displayed in Figure 4. In the VERIS recording, the P1 peak times range from 30 ms in ring one (curve 1) to 34.2 ms in ring six (curve 6). In the RETIscan-recording, the longest P1 peak time is found for ring one (curve 1, 37.5 ms) and the shortest for ring five (curve 5, 31.3 ms). Figure 5 shows boxplots of the P1 peak times of VERIS (5a) and RETIscan (5b) from our group of subjects. The main lines of the boxplot are at the 25th, 50th and 75th percentiles. The 103 hexagon pattern of VERIS leads to 6 concentric rings, the 61 hexagon pattern of RETIscan to 5 concentric rings. Despite the differences mentioned above it is obvious that the trend of the P1 peak times from the center to the periphery is different between VERIS and RETIscan. In the VERIS result, the 50th percentiles are located between 31.7 and 34.2 ms with the central peak time in ring one at 32.5 ms. The RETIscan latency-times range from 31.25 ms to 39 ms with the longest latency time in the center and shorter latencies towards the periphery. Figure 5b also shows one of the consequences of the relatively low sampling rates used with RETIscan. The curves consist of sampled points with 2.08 ms time span between 2 samples (sample times are displayed by the black horizontal lines). The reason that there are some points in the boxplot that lie between the dotted lines are changes of the sampling rates. In about 20% of the RETIscanrecordings the framerate of the stimulus monitor changed from 60 to 59 Hz resulting in a different time span between 2 samples (60 Hz: 2.08 ms, 59 Hz: 2.11 ms). This leads to slightly different time units. A more detailed discussion of the sample problem is given below (E).
B. VERIS results with an adapted setup
Figure 6a contains boxplots of the average P1 peak times for our group of subjects measured with a VERIS setup that was adapted to be as close as possible to the standard parameters of RETIscan. For example, each of the active m-frames of the m-sequence was followed by 5 f-frames. The results for the 5 rings (grouping of Figure 2, but 61 hexagon pattern) are shown. The results now show trends similar to the RETIscan recordings (Figure 5b) with the longest latency time in the center (50th percentile at 35.8 ms) and faster P1 peak times to the periphery (about 30.7 ms). In Figure 6b the averaged P1 peak times of the adapted VERIS setup and the standard RETIscan setup are displayed together. In general, the results have the same trends but with the peak times of RETIscan being slightly longer (between 0.5 and 1.7 ms) than the VERIS peak times.
169
Figure 6 (a). Boxplot of the P1 peak times of the adapted VERIS-setup.
Figure 6 (b). Averaged P1 peak times of VERIS (adapted setup) and RETIscan.
170
Figure 7 (a). Averaged P1 amplitudes for different stimulus luminance settings (N=4) recorded with VERIS.
Figure 7 (b). Averaged P1 peak times (N=4) recorded with VERIS and different stimulus luminance settings.
171 C. Influence of the different stimulus monitor settings In Figure 7a the average amplitudes of the P1 peaks of the 4 subjects and 5 concentric rings from our experiment (C) are shown. The luminance and contrast difference leads to minor changes of the amplitudes recorded (straight line: VERIS standard levels, broken line: RETIscan standard levels). The corresponding averaged P1 peak times of the 4 subjects are illustrated in Figure 7b in concentric ring analysis. In addition to the amplitude changes it can be observed that the P1 peak times of all rings are longer when the intensity difference between the black and white hexagons is bigger.
D. Influence of the m-sequence-timing In Figure 8 the topographies of the P1 peak time results of all 61 hexagons are displayed for two different m-sequence setups of VERIS. All the other parameters were exactly the same in both recordings of the same subject. With both setups the blind spot has a local maximum of the P1 peak times. With the ‘RETlscan’ setup a second maximum appears in the central hexagon. This is consistent with the concentric ring analysis of Figure 5b and 6a where the longest peak times are found in ring number 1. The peak times of the surrounding rings are getting faster with increasing distance to the center. The blind spot should be considered in a different way. It is generally thought that the response here may be caused by responses from surrounding areas which have been stimulated with reflected light from the optic nerve. Consistent with this view, the peak time of the blind spot area is significantly higher than the peak times of the surrounding hexagons in each case. If the displayed 3D-plots of Figure 8 were sectioned with a horizontal line through the central hexagon and the blind spot, the result of the left 3D-plot would mirror a single result of the boxplot in Figure 5a whereas the same section would lead to a single result of Figure 6a with the right 3D-plot in Figure 8. Of course, in both cases the value assigned to the blind spot would not appear in the boxplot because of the concentric ring analysis (average response of a complete ring). The corresponding responses (averaged to concentric rings) of each recording are displayed below the 3D topographies in normalized scale. With no f-frames (responses on the left), the concentric ring analysis of the P1 peak times shows minor changes only. At the same time, the waveforms of each ring do not change. When f-frames are inserted (responses on the right) both latencies and implicit times are varying with eccentricity. In addition, the waveform of the averaged responses has changed compared to the setup without f-frames.
172 Figure 8. P1 topographies and concentric ring grouping of the same subject (right eye) for different m-sequence-timings: original VERIS m-sequence-setup (no f-frames) on the left, adapted VERIS m-sequence-setup (with 5 f-frames) on the right. The responses of the corresponding recordings are displayed below each topography in concentric ring analysis.
173
Figure 9 (a). MERG responses of the same subject recorded with different sampling rates.
Figure 9 (b). Zoomed P1 peak of Figure 9a for different sampling rates: the maximum of the response is hit with higher sampling rates only.
174 E. Sampling rate Six subjects were measured three times with VERIS and the same parameter setup except for the sampling rate which was set to 450 Hz, 1200 Hz and 2400 Hz. Figure 9a displays three single curves of the MERG recording of one of the subjects for different sampling rates. Each of the curves consists of a number of single points. The number of points is specified by the sampling rate used, the reciprocal of which is the time gap between two samples. When the sampling rate is too low, details of the curve get lost as shown in Figure 5b in rings number two and three where all peak times seem to appear at the same point of time. The reason for the loss of information is displayed in Figure 9b where the peaks of three recordings with different sampling rates are displayed on an expanded time scale. If the time gap between two samples is too large (squared-curve, sampled with 450 Hz, time gap: 2.22 ms) it is very unlikely that the real peak of the curve is hit. If it is missed, the sampled point with the highest amplitude is described as the peak which leads to an inaccurate result. With low sampling rates the observed peak only tells you that the real peak can be found somewhere between the point before and the point after. Using higher sampling rates the problem remains, but with more points per curve the time gap between the samples is shortened and it is more likely to hit the real peak (e.g. dotted-curve in Figure 9b), and therefore the results are more reliable. The concentric ring analysis of the recordings with the different sampling rates of Figure 9 is shown in Figure 10. The curve for 450 Hz has time gaps between two sampled points of 2.22 ms. The P1 peak time differences of the real curves of rings number three, four and five should be in a region of 1 ms as can be observed in the curve with a sampling rate of 2400 Hz (timegap = 0.41 ms). These minor changes are not displayed in the 450 Hz curve because of the chosen sampling rate and, as a consequence, all of the three peak times are displayed at the same point of time (31 ms), which is not the case in reality. This example emphasizes the meaning of the sampling rate. Low sampling rates can cause severe loss of information.
Discussion The multifocal ERG responses are influenced by a combination of a number of parameters, which are both hardware and software related. When two systems are compared under clinical conditions, system-specific details will lead to limitations of the results of such comparisons. For example, to get an exact estimation of the software routines used in different systems, the stimulus monitor, amplifier and the electrode used must be the same. However, for
175
Figure 10. Concentric ring analysis for different sampling rates of a single subject: sampling rates that are too low lead to a loss of information.
clinical use, there will be no way to replace hardware equipment between two systems as only one system is available. Thus from that point of view it is more important to display the performance and the recording results of each system as it is provided by the manufacturer in order to be able to estimate the variations of the outcomes under system-specific aspects. Nevertheless it would be very interesting to repeat some of the experiments using the same hardware for both systems, for example to investigate the effects of a particular stimulus monitor or amplifier [9]. The fact that the Burian-Allen electrode could not be used with the RETIscan amplifier section also needs further investigation. In the present study, both amplitudes and peak times of recordings with different electrodes did not change significantly (Figure 2). The signal-tonoise-ratio of the Burian-Allen electrode is better than the one of the DTL, which has to be balanced using a longer recording time (e.g. twice as long). The performance of DTL and Burian-Allen electrodes has been compared in recent years, e.g. for pattern and flash ERG’s, and a reduction in response amplitude of the DTL fibers compared to the Burian-Allen was reported [10, 11]. However, the peak times measured for both DTL and Burian-Allen recordings did not show any significant differences in a recent study [11]. Thus the results of the present study should not be affected by the different
176 electrode type, no comparisons were made between amplitudes recorded with different electrodes. Moreover all of the VERIS recordings (experiment B-E and the VERIS recording of experiment A) were performed with BurianAllen electrodes only. The amplifier settings were also the same for all VERIS recordings. Here the use of an additional 50 Hz line filter has almost no effects on the recorded responses of the First-Order-Kernel, as the FOK outcomes are assembled by frequencies with main spectral components between 19 and 47 Hz [12]. Responses of higher order kernels were not taken into account because of the use of fixed short m-sequences in the standard protocol of the RETIscan system and the insertion of f-frames after each active m-sequence step (the use of f-frames leads to a severe reduction of the interactions between consecutive m-frames, which are displayed in the results of higher order kernels). If higher order kernels are to be analyzed in the future, both m-sequence length and m/f-frame combination of the m-sequence would have to be adapted. The results of experiment A illustrate the standard outcomes of VERIS and RETIscan. Valuable information can be found in the relation of the P1 peak times from the center to the periphery for each of the systems. For the RETIscan system, even with a sampling rate of 480 Hz (2.1 ms between two samples) the distance between the P1 peak times of ring 1 and ring 5 is three samples (6.3 ms) with the longest peak time in the central ring (Figure 5b). In the results recorded with the VERIS system, the P1 peak time course is different (Figure 5a). When the subjects were recorded with VERIS using an adapted setup (experiment B), the P1 peak times showed the same trend as with the RETIscan system (Figure 6a). As the results of Figure 5a and 6a have been recorded with the same hardware and software (VERIS), the changes must be caused by the parameters altered from setup A to setup B (different number of hexagons, luminance and contrast settings, m-sequence length and m-sequence timing). The effects of f-frame insertion and the use of different intensity levels of the VERIS stimulus have been demonstrated in an earlier study. It has been shown that using different stimulus intensities has effects on both amplitude and peak times of the responses. The insertion of f-frames leads to longer P1 peak times and changes of the waveforms recorded. In addition, the waveforms of the responses also vary from the center to the periphery if f-frames are used [3]. Those findings are confirmed by experiments C and D even though if 5 f-frames have been used between two active m-steps. The use of different stimulus luminance levels leads to minor changes of the response amplitudes (Figure 7a). At the same time, brighter stimuli lead to longer P1 peak times (Figure 7b). The waveform changes caused by the insertion of f-frames are il-
177 lustrated in Figure 8. Without f-frames (left side), in general the waveforms of the responses do not change with eccentricity in the concentric ring analysis. Using f-frames, waveform changes can be observed which can be described most effectively with the performance of the P1 peak where the longest latencies are found in the central region. Thus the topography of the P1 peak times is significantly different (Figure 8). In this context, the effects of the contrast changes from the flicker stimulus (1:1 with no f-frames, 0.5:5.5 with 1m/5f frames) to the recorded responses need further investigation as adaptation mechanisms begin to have major impact on the recording results [3]. Other parameters also have impact on the recorded data, for example the sampling rate, which primarily is not a basic parameter of the multifocal ERG. Thus the sampling rate has to be selected according to the target of the corresponding recordings. In general, using a higher sampling rate is the better choice. On the other hand, the data size of the recordings gets bigger at the same time, so the files will need more disk space and more time to get processed when the recording is loaded. When the recording protocols of VERIS and RETIscan are close, the results show some comparable trends, despite the remaining differences of the stimulus and the hardware used (Figure 6b). However, in clinical use, the differences between those results will grow because the aim of the investigation and the signal-to-noise-situation of the environment will force the user to draw out the best of each system, resulting in parameter changes which may lead in different directions. Most importantly one has to keep in mind that the m-sequence-timing has a large effect on the results and that secondary parameters such as the sampling rate should be chosen carefully in order not to lose valuable information.
Acknowledgments Supported by the DFG (Lo457/3-1). We thank Dr. Sutter for his help with our first steps with VERIS, Dr. Hood for his useful comments with the manuscript and his help with our first steps in data analysis, and RolandConsult for giving us the opportunity to test their system.
Note Part of this work has been presented as a poster at the 36th Symposium of the International Society for Clinical Electrophysiology of Vision 1998 in Hradec Kralove, Czech Republic.
178 References 1. 2. 3. 4. 5.
6. 7. 8.
9. 10. 11. 12.
Sutter EE, Tran D. The field topography of ERG components in man–l. The photopic luminance response. Vision Research 1992; 32: 433–46. Bearse MA, Sutter EE. Imaging localized retinal dysfunction with the multifocal electroretinogram. J Opt Soc Am A 1996; 13: 634–40. Hood DC, Seiple W, Holopigian K, Greenstein V. A comparison of the components of the multifocal and full-field ERGs. Vis Neurosci 1997; 14: 533–44. Hood DC, Li J. A technique for measuring individual multifocal ERG records. In: Yager D, eds. Non-invasive assessment of the visual system. Opt Soc Am 1997; 11: 33–41. Hood DC, Holopigian K, Greenstein V, Seiple W, Li J, Sutter EE, Carr RE. Assessment of local retinal function in patients with retinitis pigmentosa using the multi-focal ERG technique. Vision Res 1998; 38: 163–79. Seeliger M, Kretschmann U, Apfelstedt-Sylla E, Zrenner E. Implicit Time Topography of Multifocal Electroretinograms. Vision Res 1996; 36: 64. Seeliger M, Kretschmann U, Apfelstedt-Sylla E, Ruther K, Zrenner E. Multifocal electroretinography in retinitis pigmentosa. Am J Ophthalmol 1998; 125: 214–26. Kretschmann U, Seeliger M, Ruther K, Usui T, Apfelstedt-Sylla E, Zrenner E. Multifocal electroretinography in patients with Stargardt’s macular dystrophy. Br J Ophthal 1998; 14: 267–75. Keating D, Parks S, Evans AL, Williamson TH, Elliot AT, Jay U. The effect of filter bandwidth on the multifocal electroretinogram. Doc Ophthalmol 1997; 92: 291–300. Hennessy MP, Veagan. Amplitude scaling relationships of Burian-Allen, gold foil and Dawson, Trick and Litzkow electrodes. Doc Ophthalmol 1995; 89: 235–48. McCulloch DL, Van Boemel GB, Borchert MS. Comparisons of contact lens, foil, fiber and skin electrodes for pattern electroretinograms. Doc Ophthalmol 1998; 94: 327–40. Bock M, Andrassi M, Gerth C, Lorenz B. Spectral mapping of FOK and SOK responses from multifocal ERGs. Proceedings of the 37th ISCEV Meeting, Eilat, 11.4.-16.4.1999, ISCEV abstract book 1999; 15.
Address for correspondence: M. Bock, Department of Paediatric Ophthalmology, Strabismology and Ophthalmogenetics, University of Regensburg, Franz-Josef-Strauß-Allee 11, 93053 Regensburg, Germany Phone: +49-941 -9449229; Fax: +49-941-9449216; E-mail:
[email protected]
