Normalization of Visual Evoked Potentials Using Underlying Electroencephalogram Levels Improves Amplitude Reproducibility in Rats

Home , Evoked potential

Visual Neurophysiology Normalization of Visual Evoked Potentials Using Underlying Electroencephalogram Levels Improves Amplitude Reproducibility in Rats

Yuyi You,1 Johnson Thie,1 Alexander Klistorner,1,2 Vivek K. Gupta,1 and Stuart L. Graham1,2

PURPOSE. The visual evoked potential (VEP) is a frequently used encephalogram (EEG)-based signals that form the family of noninvasive measurement of visual function. However, high- cortical evoked potentials (EPs).2 The ongoing cortical activity, amplitude variability has limited its potential for evaluating which has a major influence on sensory processing, should not axonal damage in both laboratory and clinical research. This be regarded as merely random background noise.3–5 In a neu- study was conducted to improve the reliability of VEP ampli- rophysiologically based mathematical model, Jansen et al.6 tude measurement in rats by using electroencephalogram demonstrated that the spontaneous EEG and the flash VEP are (EEG)-based signal correction. generated by some of the same neural structures. METHODS. VEPs of Sprague-Dawley rats were recorded on three It was suggested that the amplitude of the VEPs reflects the separate days within 2 weeks. The original VEP traces were number of functional optic nerve fibers. Our previous work re- normalized by EEG power spectrum, which was evaluated by vealed a positive correlation between retinal nerve fiber layer 7,8 Fourier transform. A comparison of intersession reproducibil- thickness and VEP amplitude in optic neuritis patients. We ity and intersubject variability was made between the original recently also demonstrated a strong linear relationship between and corrected signals. the axonal loss and amplitude decrease in experimental optic nerve demyelination.9 However, the relationship between axonal RESULTS. Corrected VEPs showed lower amplitude intersession damage and amplitude was weaker than that between demyeli- within-subject SD (Sw), coefficient of variation (CoV), and nation and latency delay, which may be due to the higher vari- repeatability (R ) than the original signals (P Ͻ 0.001). The 95 ability of the VEP amplitude.9 The VEP has been extensively used intraclass correlation coefficient (ICC) of the corrected traces in pharmacologic experiments and in assessing the neuroprotec- (0.90) was also better than the original potentials (0.82). For tive effects of compounds, as it provides a means of monitoring intersubject variability, the EEG-based normalization improved neural activity and sensory processing in vivo.10 The reproduc- the CoV from 44.64% to 30.26%. A linear correlation was ibility of VEP amplitude is critical if the parameter is used to observed between the EEG level and the VEP amplitude (r ϭ evaluate nerve functional change or neuronal apoptosis over time 0.71, P Ͻ 0.0001). in some optic nerve disease models.11 CONCLUSIONS. Underlying EEG signals should be considered in Rats are the most common species used in laboratory VEP measuring the VEP amplitude. In this study, a useful technique studies because of their accessibility and short growth cycle.12 was developed for VEP data processing that could also be used We have recently improved the reproducibility of VEP mea- for other cortical evoked potential recordings and for clinical surements in rats by using a modified electrode configuration, VEP interpretation in humans. (Invest Ophthalmol Vis Sci. a novel mini-Ganzfeld visual stimulator, and early-peak analy- 2012;53:1473–1478) DOI:10.1167/iovs.11-8797 sis.13 However, the variability of the amplitude was still rela- tively high, which may limit the use of the VEP in assessing he visual evoked potential (VEP) is a measurement in visual axonal damage. The variability of VEP amplitude has been Telectrophysiology that is frequently used as an objective attributed to many factors in both animals and humans, such as means of evaluating the function of the visual pathway. It is cortical anatomy,14 the conductivity of the electrodes and understood to be generated at the level of the striate cortex by underlying tissues,13 strain and age,15 stimulus design,13,16 and 1 combined activity of postsynaptic potentials. The VEP is an general level of cortical activity.17–19 Some factors (e.g., elec- analogous phenomenon of auditory evoked potential (AEP) trode configuration and stimulus intensity) are easy to rule out and somatosensory evoked potential (SEP), and all are electro- by using a standardized recording protocol. However, the ongoing cortical activity is very difficult to control during different recording sessions. As mentioned, the VEP is an EEG- 1 based signal, and we have reported the effect of EEG on From the Department of Ophthalmology, Australian School of 20 Advanced Medicine, Macquarie University, Sydney, New South Wales, multifocal VEP in humans. Australia; and the 2Save Sight Institute, Sydney University, Sydney, New In this study, we scaled the original VEP traces using the South Wales, Australia. ongoing EEG power spectrum level to minimize the effect of Supported by the Ophthalmic Research Institute of Australia brain activity and to improve the reliability of VEP amplitude (ORIA) and the National Multiple Sclerosis Society (NMSS). AK was measurement, as just described. We repeatedly performed the supported by the Sydney Foundation for Medical Research. VEP recording in rats by using a well-established protocol9,13 Submitted for publication October 14, 2011; accepted January 23, and analyzed the intersession reproducibility and intersubject 2012. variability of the original and corrected VEPs, respectively. The Disclosure: Y. You, None; J. Thie, None; A. Klistorner, None; V.K. Gupta, None; S.L. Graham, None purpose of this study was to explore a reliable data-processing Corresponding author: Yuyi You, Australian School of Advanced technique to improve VEP amplitude reproducibility for labo- Medicine, Macquarie University, F10A, Level 1, 2 Technology Pl, North ratory research in animal models. The technique could also be Ryde, Sydney, NSW 2109, Australia; [email protected]. considered for clinical VEP interpretation in humans.

Downloaded from tvst.arvojournals.org on 09/24/2021 1474 You et al. IOVS, March 2012, Vol. 53, No. 3

METHODS reference screw were connected to the amplifier using hook clips. A needle electrode (F-E3M-72; Grass, Warwick, RI) was inserted into the Animals tail as the ground. The electrode impedance was measured (F-EZM5 Impedance Meter; Grass) and maintained below 5 k⍀. Visual stimuli Twenty-five male Sprague-Dawley rats with a body weight of 300 to were generated by a mini-Ganzfeld stimulator (3 cd ⅐ s/m2), and the 350 g (10–12 weeks; Animal Research Centre, Perth, Western Austra- photic stimulation was delivered 100 times at a frequency of 1 Hz. The lia, Australia) were used. All animals were maintained in an air-condi- time of measurement was 200 ms, and the sampling rate was 5000 tioned room with controlled temperature (21 Ϯ 2 °C) and fixed daily samples/s. Responses were amplified 20,000 times and filtered by 12-hour light/dark cycles. All procedures involving animals were con- band-pass filter, with low and high cutoff frequencies of 1 and 100 Hz, ducted in accordance with the Australian Code of Practice for the Care respectively (BMA-400 Bioamplifier, CWE, Inc., Ardmore, PA). After and Use of Animals for Scientific Purposes and the guidelines of the the recording, the wound was closed and antibiotics administered. ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. The animals were anesthetized with an intraperitoneal in- EEG-Based Correction jection of ketamine (75 mg/kg) and medetomidine (0.5 mg/kg) for both surgery and electrophysiology recording. Fourier transform was used to quantify the amplitude of the underlying EEG during each recording session. The Fourier coefficients of the EEG VEP Recording sequence, x[n], were evaluated on the entire 100-second sequence. Let X[m] denote the Fourier coefficient, defined as Stainless steel skull screws (Micro Fasteners, Melbourne, Australia) were implanted through the skull into the visual cortex (7 mm behind the NϪ1 bregma and 3 mm lateral of the midline) as the positive electrodes. A 1 X͓m͔ ϭ ͸x͓n͔eϪ͓͑2␲imn͒/M͔, m ϭ 0,...,M Ϫ 1. reference screw electrode was placed on the midline, 3 mm rostral to the N bregma. Dental cement (Rapid Repair; DeguDent GmbH, Hanau, Ger- nϭ0 many) was used to fix the screws. The skin of the head was closed, and at least 1 week was allowed for the animals to recover from the surgery. The Fourier transform was performed by MatLab Function FFT with 13 VEP recording was performed on the same eye (randomly selected) M ϭ 2 (The MathWorks, Inc., Natick MA). The choice of M allows the on three separate days within 2 weeks. The animals were anesthetized, use of the Fast Fourier transform algorithm and ensures that the placed in a dark room, and allowed to adapt to darkness for 5 minutes frequency interval is at most 1 Hz/point. (The sampling frequency was (we have recently confirmed the adequacy of such a short adaptation 5000 samples per second.) time using our methodology).9,13 The body temperature was main- The amplitude of EEG was evaluated according to 30-Hz band tained at 37°C by a homoeothermic blanket system with a rectal power thermometer probe (Harvard Apparatus; Holliston, MA). The pupils b were dilated with 1.0% tropicamide eyedrops (Alcon Laboratories, Fort 1 Worth, TX). The skin over the skull was opened. The positive screw EEG ϭ ͸X2͓m͔ b Ϫ a ϩ 1 over the contralateral visual cortex of the stimulated eye and the mϭa

FIGURE 1. (A) Raw EEG sequence (100 seconds). (B) Corresponding EEG power spectrum (Fast Fourier transform, 1–30 Hz).

Downloaded from tvst.arvojournals.org on 09/24/2021 IOVS, March 2012, Vol. 53, No. 3 EEG-Based VEP Normalization 1475

FIGURE 2. (A) Underlying EEG signals from an individual rats on three separate days. (B) Intersession original VEP traces. (C) Corrected VEP traces based on the corresponding EEG signals.

where a and b are the indexes that correspond to 1 and 30 Hz, RESULTS respectively (Fig. 1). The VEP traces were then corrected by the following formula: Intersession Reproducibility

ϭ ϫ ͑ ͒ Figure 2 presents the original and corrected VEP traces from an VEPC VEPO EEGAVER/EEGIND individual rat and the corresponding EEG signals on 3 days. The original VEP amplitude on day 2 was smaller than those on days 1 and 3 (Fig. 2B), and the corresponding EEG amplitude on day where VEPC and VEPO represent corrected VEP and original VEP 2 was also the smallest (Fig. 2A). The intersession reproduc- traces, respectively; EEGAVER is the average of EEGs from all the ibility parameters are summarized in Table 1. The rectified subjects; and EEGIND is the EEG recorded from an individual rat in a particular recording session. VEPs showed significant lower amplitude Sw, CoV, and R95 than the original VEPs. The CoV value improved from 24.09% (original) to 16.49% (corrected). The intersession ICCs were Data Analysis 0.82 for the original and 0.90 for the corrected signals, respec- The largest peak-to-trough amplitudes20 for each VEP wave within the tively. initial 100-ms interval were determined and used for amplitude analysis. Intersession reproducibility (recordings on three separate days) Intersubject Variability was determined by calculating the intersession within-subject SD (Sw), 21,22 coefficient of variation (CoV), repeatability (R95), and intraclass Figure 3 shows the representative VEP traces from two indi- correlation coefficient (ICC).23 The day 1 data were also used to vidual rats before and after amplitude correction. The final VEP evaluate intersubject variability, which was determined by between- amplitudes of these two rats became similar after EEG-based subject SD (Sb) and intersubject CoV. The comparison of intersession normalization (Fig. 3C). The amplitudes of original and cor- reproducibility between original and corrected VEP signals was made rected VEPs from all the rats are plotted in Figure 4. The by paired t-test. The F-test was used for the between-subject standard amplitude Sb was 21.21 ␮V for the raw traces and only 14.12 deviations. The correlation between EEG and VEP amplitudes was also ␮V for the rectified VEPs (P ϭ 0.02). Similarly, the intersubject examined by linear regression analysis. Statistical significance was CoV decreased from 44.64% to 30.26% after EEG-based correc- defined as P Ͻ 0.05 (Prism, ver. 5.0; GraphPad, San Diego, CA). tion.

TABLE 1. Comparison of Sw, CoV, R95, and ICC between Original and EEG-Corrected VEP Amplitudes ␮ ␮ Sw ( V) CoV (%) R95 ( V) ICC

Original 10.28 (6.25–14.30) 24.09 (14.65–33.53) 28.46 (17.31–39.62) 0.82 (0.65–0.92) Corrected 7.31 (4.44–10.17) 16.49 (10.02–22.95) 20.24 (12.30–28.17) 0.90 (0.80–0.95) P 0.001* Ͻ0.001* 0.001* —

n ϭ 25; data are the mean (95% CI). *P Ͻ 0.05 (paired t-test).

Downloaded from tvst.arvojournals.org on 09/24/2021 1476 You et al. IOVS, March 2012, Vol. 53, No. 3

FIGURE 3. (A) Underlying EEGs from two individual rats. (B) Original VEP traces from these two rats. (C) VEP traces after EEG-based correction.

Correlation between EEG Level ical signals need to be compared to detect changes in optic and VEP Amplitude nerve fibers. The generation of VEPs is related to both visual stimulation The linear regression analysis between VEP amplitude and the and intrinsic activity in the neuronal networks of the brain. The ongoing EEG power spectrum level is shown in Figure 5. A ϭ Ͻ intrinsic activity has a larger influence on cortical activity fairly high r value (r 0.71, P 0.0001) was observed patterns than visual stimuli, even in the primary visual cortex.4 between the amplitudes of these two cortical potentials. Therefore, the amplitude of VEP is rather small compared with that of the ongoing EEG. The external stimulation has to be repeated, usually 100 times, to minimize the noise from EEG, DISCUSSION and it is estimated that after 100 repetitions, the signal-to-noise 2 The VEP has been used to assess visual performance in humans ratio is increased 10-fold. The effect of sleep–awake states on 19 and in a wide variety of animals.12 VEP recording in rats has rat VEP waveforms has been established by Meeren et al. The been well described in studying the visual plasticity and in normalization of the original VEP signal in relation to the evaluating the visual system integrity.11,24–26 Recently, we underlying EEG can compensate for the effect of higher corti- have optimized the VEP recording protocol and designed a cal activity on VEP amplitude. In animal research, different mini-Ganzfeld flash stimulator13 that provided superior eye types of anesthetics and the level of anesthesia can affect the isolation to assess the optic nerve conduction in an optic measurement of VEP parameters.27,28 This effect is also poten- neuritis model.9 In the present study, we developed an EEG- tially minimized using the EEG-based correction, as the back- based signal processing technique, seeking to improve the ground brain activity is standardized. reproducibility of VEP amplitude measurement in rats. It was The conductivity of the electrodes and the underlying tissue suggested that the amplitude of VEP reflects the axonal damage also has significantly impacts on the VEP amplitude. In rat in the visual pathway. Therefore, the intersession reproducibil- recording with implanted screws, a major factor is the connec- ity is very important in VEP experiments, because chronolog- tion between hook clips and skull screw electrodes, which can

Downloaded from tvst.arvojournals.org on 09/24/2021 IOVS, March 2012, Vol. 53, No. 3 EEG-Based VEP Normalization 1477

FIGURE 6. Schematic circuit diagram representing the effect of elec-

trode conductivity on evoked potential amplitude. RB, the internal resistance of the rat’s brain between positive and negative screws; RCS, FIGURE 4. Scatterplot of original and corrected VEP amplitudes. The the resistance of skull screws plus the resistance of the joint between amplitudes from original VEPs are more variable. Error bars, SEM. screws and hook clips (ideally 0); RC, the resistance of hook clips and wires; and RCC, the external resistance of the rat’s skull between hook clips. RC is very small and is approximately 0, because the hook clips ⍀Ϸ ϩ ϩ ϫ be affected by adjacent blood clots and surrounding connec- are made of copper. The total impedance is (RB RCS1 RCS2) ϩ ϩ ϩ Ͻ ⍀ tive tissue edema. The impedance between the positive and RCC/(RB RCS1 RCS2 RCC) 5k . According to Ohm’s law, the ϭ ϫ ϩ ϩ ϩ negative electrodes should be routinely measured and con- measured EP amplitude VEP V RCC/(RB RCS1 RCS2 RCC), trolled below 5 k⍀ (usually, 2–3 k⍀).29 However, when the where V is the real potential in the brain between positive and negative electrodes. Under an ideal recording condition, R , R , R ϽϽ R , resistance of the interface between screws and hook clips in- B CS1 CS2 CC І V Ϸ V. creases, there is less current flow (I) through the amplifier, and EP thereby the input voltage to the amplifier is reduced (Fig. 6). Therefore, a poor electrode connection leads to smaller VEP necessary to include a large sample to achieve statistical signif- amplitudes. In theory, using the underlying EEG-based correction icance (e.g., P Ͻ 0.05). Our current results showed that after eliminates this electrode conductivity factor. the EEG-based correction the intersubject variability also sig- Indeed, in the present study, we observed a significant nificantly improved (Figs. 3, 4), and the intersubject CoV de- improvement in the intersession reproducibility of the VEP creased from 44.64% to 30.26%. The intersubject variability in amplitude after normalizing the raw signals (Fig. 2; Table 1). VEP may also be due to the various levels of general brain The CoV dropped from approximately 25% to 15%. This VEP activity and the electrode conductivity as described above, analysis technique would be very useful in visual electrophys- which can be standardized using the EEG-based correction. iological experiments, as it improves the reliability of data However, some other factors, such as age15,30,31 and individual interpretation. However, the CoV of the amplitude after EEG- cortical anatomy,12,14 that also contribute to intersubject vari- based correction was still higher than that of the latency Ϸ 13 ability cannot be eliminated. The positive linear correlation (CoV 5%) from a previous study, which indicates that (Fig. 5) between the EEG and VEP amplitudes observed in this some other cortical activity aspects (e.g., different rhythms of study was in accordance with the results from some previous the EEG) may also play a role in affecting the amplitude mea- studies of evoked potentials in both animals and humans,4,18,20 surement. which supported the assumption that similar influences ac- The higher intersubject variability of VEP amplitude has count for amplitude variability of these two cortical poten- limited its use in electrophysiological research, as it may be tials.20 In conclusion, we developed a practical technique for VEP data processing and interpretation. We observed a strong correlation between the VEP and the ongoing EEG amplitudes and significantly improved the intersession reproducibility and intersubject variability of the VEP amplitude measurement. We highly recommend that underlying EEG signals be considered in measuring the VEP amplitude. EEG-based VEP amplitude correction would be very useful, not only for laboratory research, but also for clinical VEP interpretation. Furthermore, as the VEP is a cortical evoked potential analogous to AEP and SEP, we hypothesize that this technique has potential for use in improving the reliability of other EP recordings.

References 1. Creutzfeldt O, Maekawa K, Hosli L. Forms of spontaneous and evoked postsynaptic potentials of cortical nerve cells. Prog Brain Res. 1969;31:265–273. 2. Fahle M, Bach M. Origin of the visual evoked potentials. In: Heck- FIGURE 5. Linear regression analysis between original VEP ampli- enlively J, Arden G, eds. Principles and Practice of Clinical Elec- tudes and corresponding EEG levels (Fourier transform power spec- trophysiology of Vision. Cambridge, MA: MIT Press; 2006:207– trum; r ϭ 0.71, P Ͻ 0.0001). 234.

Downloaded from tvst.arvojournals.org on 09/24/2021 1478 You et al. IOVS, March 2012, Vol. 53, No. 3

3. Arieli A, Shoham D, Hildesheim R, Grinvald A. Coherent spatio- 17. Basar E, Rahn E, Demiralp T, Schurmann M. Spontaneous EEG temporal pattern of on-going activity revealed by real-time optical theta activity controls frontal visual evoked potential amplitudes. imaging coupled with single unit recording in the cat visual cortex. Electroencephalogr Clin Neurophysiol. 1998;108:101–109. J Neurophysiol. 1995;73:2072–2093. 18. Brandt M, Jansen B. The relationship between prestimulus-alpha 4. Arieli A, Sterkin A, Grinvald A, Aertsen A. Dynamics of ongoing amplitude and visual evoked potential amplitude. Int J Neurosci. activity: explanation of the large variability in evoked cortical 1991;61:261–268. potentials. Science. 1996;273:1868–1871. 19. Meeren H, Van Luijtelaar E, Coenen A. Cortical and thalamic visual 5. Fellinger R, Klimesch W, Gruber W, Freunberger R, Doppelmayr evoked potentials during sleep-wake states and spike-wave dis- M. Pre-stimulus alpha phase-alignment predicts p1-amplitude. charges in the rat. Electroencephalogr Clin Neurophysiol. 1998; Brain Res Bull. 2011;85:417–423. 108:306–319. 6. Jansen B, Zouridakis G, Brandt M. A neurophysiologically based 20. Klistorner A, Graham S. Electroencephalogram-based scaling of mathematical model of flash visual evoked potentials. Biol Cybern. multifocal visual evoked potentials: effect on intersubject ampli- 1993;68:275–283. tude variability. Invest Ophthalmol Vis Sci. 2001;42:2145–2152. 7. Klistorner A, Arvind H, Nguyen T, et al. Axonal loss and myelin in 21. Bland J, Altman D. Measurement error. BMJ. 1996;313:744–753. early ON loss in postacute optic neuritis. Ann Neurol. 2008;64: 22. Bland J, Altman D. Statistical methods for assessing agreement 325–331. between two methods of clinical measurement. Lancet. 1986;1: 8. Klistorner A, Arvind H, Garrick R, Graham S, Paine M, Yiannikas C. 307–310. Inter-relationship of optic coherence tomography and multifocal 23. McGraw K, Wong S. Forming inferences about some intraclass visual-evoked potentials after optic neuritis. Invest Ophthalmol correlation coefficients. Psychol Methods. 1996;1:30–46. Vis Sci. 2010;51:2770–2777. 24. Miyake K-I, Yoshida M, Inoue Y, Hata Y. Neuroprotective effect of 9. You Y, Klistorner A, Thie J, Graham S. Latency delay of visual transcorneal electrical stimulation on the acute phase of optic evoked potential is a real measurement of demyelination in a rat nerve injury. Invest Ophthalmol Vis Sci. 2007;48:2356–2361. model of optic neuritis. Invest Ophthalmol Vis Sci. 2011;52:6911– 25. Jehle T, Wingert K, Dimitriu C, et al. Quantification of ischemic 6918. damage in the rat retina: a comparative study using evoked poten- 10. Iwamura Y, Fujii Y, Kamei C. The effects of certain H1-antagonists tials, electroretinography, and histology. Invest Ophthalmol Vis on visual evoked potential in rats. Brain Res Bull. 2003;61:393– Sci. 2008;49:1056–1064. 398. 26. Guire E, Lickey M, Gordon B. Critical period for the monocular 11. Meyer R, Weissert R, Diem R, et al. Acute neuronal apoptosis in a deprivation effect in rats: assessment with sweep visually evoked rat model of multiple sclerosis. J Neurosci. 2001;21:6214–6220. potentials. J Neurophysiol. 1999;81:121–128. 12. Ridder W. Visual evoked potentials in animals. In: Heckenlively JR, 27. Henry K, Rhoades R. Relation of albinism and drugs to the visual Arden GB, eds. Principles and Practice of Clinical Electrophysi- evoked potential of the mouse. J Comp Physiol Psychol. 1978;92: ology of Vision. Cambridge, MA: The MIT Press; 2006:935–947. 271–279. 13. You Y, Klistorner A, Thie J, Graham S. Improving reproducibility 28. Castro-Junior J, Resende L, Bertotti M, Fonseca R, Zanchetta S, of VEP recording in rats: electrodes, stimulus source and peak Schelp A. Comparison of the effects of barbiturate, benzodiazepine analysis. Doc Ophthalmol. 2011;123:109–119. and ketamine on visual evoked potentials in rabbits. Electromyogr 14. Baseler H, Sutter E, Klein S, Carney T. The topography of visual Clin Neurophysiol. 2005;45:259–262. evoked response properties across the visual field. Electroen- 29. Odom JV, Bach M, Brigell M, et al. ISCEV standard for clinical visual cephalogr Clin Neurophysiol. 1994;90:65–81. evoked potentials (2009 update). Doc Ophthalmol. 2010;120:111– 15. Heiduschka P, Schraermeyer U. Comparison of visual function in 119. pigmented and albino rats by electroretinography and visual 30. Allison T, Hume A, Wood C, Goff W. Developmental and aging evoked potentials. Graefes Arch Clin Exp Ophthalmol. 2008;246: changes in somatosensory, auditory and visual evoked potentials. 1559–1573. Electroencephalogr Clin Neurophysiol. 1984;58:14–24. 16. Creel D, Dustman RE, Beck EC. Intensity of flash illumination and 31. Celesia G, Kaufman D, Cone S. Effects of age and sex on pattern the visually evoked potential of rats, guinea pigs and cats. Vision electroretinograms and visual evoked potentials. Electroencepha- Res. 1974;14:725–729. logr Clin Neurophysiol. 1987;68:161–171.

Downloaded from tvst.arvojournals.org on 09/24/2021