A, The multifocal visual-evokedpotential (mfVEP) stimulus display: The dartboard pattern consists of 60 sectors,each with a checkerboard pattern of 16 checks, 8 white (200 candelas [cd]/m2) and 8 black (<3 cd/m2). The sectors are corticallyscaled with eccentricity, and the entire display has a diameter of 44.5°.B, An mfVEP record showing the signal and noise windows used for analysis.
A, Monocular multifocal visual-evokedpotential (mfVEP) probability plot for the left eye of a patient. B, An mfVEPprobability plot for the right eye of the same patient. C, An example of aninterocular probability plot. The blue squares indicate locations with signal-noiseratio (SNR) values that fell more than 1.96 (light blue) or 2.58 (dark blue)SDs below the mean values. The gray square indicates a location where responsesfrom both eyes have an SNR of less than 1.7. D, The mfVEP responses obtainedfrom the left (red) and right (blue) eyes of a patient. The responses withinthe ellipse are significantly smaller in the right eye.
The visual field (A), the opticnerve head (B), and the multifocal visual-evoked potential probability plot(C) divided into 6 sectors.
All data are for the same patient.A and B, Humphrey visual field total deviation probability plots for the leftand right eyes, respectively. The dark red squares (A) indicate those locationswith P<.01 vs the age-matched healthy controlsubjects; and the light blue square (B), the location with P<.05 vs the control subjects. C and D, Confocal scanning laserophthalmoscopic (Heidelberg Retina Tomograph II) rim sector maps for the leftand right eyes, respectively. The red circles (C) indicate that rim area sectors4, 5, and 6 were classified as being outside the normal limits. E and F, Monocularmultifocal visual-evoked potential (mfVEP) probability plots for the leftand right eyes, respectively. G, An interocular mfVEP probability plot. ForE-G, the black squares indicate that there was no significant difference vscontrol subjects; and the colored squares (red for the left eye and blue forthe right eye), locations with values that fell more than 1.96 (light color)or 2.58 (dark color) SDs below the mean values.
All data are for the same patient.A and B, Humphrey visual field total deviation probability plots for the leftand right eyes, respectively. In B, the dark blue squares indicate those locationswith P<.01 vs the age-matched healthy controlsubjects; and the light blue squares, those locations with P<.05 vs the control subjects. C and D, Confocal scanning laserophthalmoscopic (Heidelberg Retina Tomograph II) rim sector maps for the leftand right eyes, respectively. E and F, Monocular multifocal visual-evokedpotential (mfVEP) probability plots for the left and right eyes, respectively.G, An interocular mfVEP probability plot. For E-G, the black squares indicatethat there was no significant difference vs control subjects; and the coloredsquares (red for the left eye and blue for the right eye), locations withvalues that fell more than 1.96 (light color) or 2.58 (dark color) SDs belowthe mean values.
The locations of Humphrey visualfield (HVF), multifocal visual-evoked potential (mfVEP), and rim area sector,or confocal scanning laser ophthalmoscopic (Heidelberg Retina Tomograph II[HRT]), defects for 20 patients (A-T). For each patient, grids on the leftindicate data for left eyes; and grids on the right, data for right eyes.A hatched square indicates that the sector had a significant defect.
Greenstein VC, Thienprasiddhi P, Ritch R, Liebmann JM, Hood DC. A Method for Comparing Electrophysiological, Psychophysical, and StructuralMeasures of Glaucomatous Damage. Arch Ophthalmol. 2004;122(9):1276-1284. doi:10.1001/archopht.122.9.1276
Copyright 2004 American Medical Association. All Rights Reserved.Applicable FARS/DFARS Restrictions Apply to Government Use.2004
To develop a method for comparing multifocal visual-evoked potential(mfVEP) responses and behaviorally determined visual fields with structuralmeasures of the optic nerve head.
Humphrey 24-2 visual fields and mfVEPs were obtained from each eye of20 patients with open-angle glaucoma. Monocular and interocular analyses wereperformed to identify locations with abnormal mfVEP responses. Optic discswere assessed with a confocal scanning laser ophthalmoscope (Heidelberg RetinaTomograph II). The image of the optic nerve head was divided into 6 sectors.The rim and disc area measurements for each sector were compared with thosein a normal database using Moorfields regression analysis. The optic nervehead measurements for the 6 sectors were related to the Humphrey visual fieldlocations and the 60 sectors of the mfVEP display.
Of 240 sectors tested (40 eyes × 6 sectors), 18.8% on Humphreyvisual field, 22.1% on mfVEP, and 10.8% on confocal scanning laser ophthalmoscopictesting were significantly different from those of control subjects. Therewere no significant deficits in 165 sectors. There was agreement for 86.7%of the sectors when the Humphrey visual field and mfVEP results were compared.The confocal scanning laser ophthalmoscopic results were in agreement for84.6% of these sectors.
The method used allows for a comparison among measures of visual functionand a structural measure of the optic nerve head. In general, the resultsof the functional and structural measures showed agreement; however, therewere clear examples of disagreements that merit further study.
Glaucoma is a progressive optic neuropathy in which loss of retinalganglion cells leads to a characteristic pattern of optic nerve head and visualfield damage. The importance of detecting the presence and extent of the structuraland functional damage cannot be overestimated in the treatment of the patientwith glaucoma. Structural deficits can be evaluated quantitatively with fundalimaging techniques such as the confocal scanning laser ophthalmoscope, andfunctional deficits can be evaluated with static automated achromatic perimetry.Although the latter has become the clinical standard for detecting and monitoringglaucomatous visual field deficits, there are problems with this technique.For some patients, it is difficult, or even impossible, to obtain reliablevisual field measures. In addition, a significant loss of ganglion cells canoccur before the development of visual field loss.1- 4
A new technique, the multifocal visual-evoked potential (mfVEP), hasgenerated considerable interest as a potential solution to these problems.Studies5- 11 haveshown that the technique can detect local damage. However, the extent to whichmfVEP results will augment information obtained with static automated perimetryremains to be determined.12
The problems of obtaining reliable and repeatable measurements are notunique to techniques that assess functional damage. Techniques that assessstructural damage (eg, confocal scanning laser ophthalmoscopy) face similarproblems of variability and test-retest reliability. It is only by comparingthe results of measures of structural change with those of functional changethat our ability to diagnose and monitor glaucoma will be improved.
Here, we provide a method for comparing the results of functional andstructural measures in a group of patients with open-angle glaucoma. The aimis to develop an easy way for displaying and comparing the results of testssupplying topographic information about glaucomatous damage. The results of2 techniques for assessing functional damage, standard automated achromaticvisual fields and the mfVEP, are compared with the results of a techniquefor assessing optic disc damage, confocal scanning laser ophthalmoscopy.
Twenty white patients, aged 40 to 74 years (mean ± SD age, 59.0± 8.7 years), with open-angle glaucoma were enrolled. All patientshad glaucomatous optic nerve damage in at least 1 eye, defined as a cup-discratio of 0.6 or more, cup-disc asymmetry between fellow eyes of greater than0.2, rim thinning, notching, excavation, and/or retinal nerve fiber layerdefects and achromatic visual field loss. No specific intraocular pressurewas required to diagnose open-angle glaucoma. Of the 20 patients, 9 had primaryopen-angle glaucoma, 10 had normal-tension glaucoma, and 1 had pigmentaryglaucoma. Inclusion criteria were a visual acuity of 20/25 or better (19 patientshad a visual acuity of 20/20, and 1 had a visual acuity of 20/25), refractiveerrors not exceeding 5.00 diopters sphere or 2.00 diopters cylinder, undilatedpupillary diameters of at least 2.5 mm, reproducible visual field defects,an open anterior chamber angle, and no other posterior segment eye disease.
Thirty control subjects, aged 20 to 62 years (mean ± SD age,36 ± 13 years), with no history or evidence of ocular disease and acorrected visual acuity of 20/20 or better also participated. All subjectsgave written informed consent before participating in the study. The protocolwas approved by the committee of the Institutional Board of Research Associatesof Columbia University, and procedures followed the tenets of the Declarationof Helsinki.
Standard full-threshold or SITA (Swedish interactive threshold algorithm)-standardautomated achromatic perimetry was performed with an analyzer (Humphrey FieldAnalyzer II; Humphrey Systems, Inc, Dublin, Calif) using program 24-2. Botheyes were tested. The mean (median) value of the mean deviation of the Humphreyvisual field (HVF) results for this group of patients was −2.72 (−2.72)dB; the range was from 1.56 to −7.84 dB. All patients had prior experiencewith automated perimetry using the analyzer, and they all had reliable fieldsin both eyes as determined by the HVF statistics. Reliable visual fields weredefined as those with fewer than 33% fixation losses, false-positive results,and false-negative results.
The mfVEP testing was performed on both eyes using computer software(VERIS; Electro-Diagnostic Imaging, San Mateo, Calif). The stimulus is shownin Figure 1A. The dartboard patternconsisted of 60 sectors, each with a checkerboard pattern of 16 checks, 8white (200 candelas [cd]/m2) and 8 black (<3 cd/m2).The sectors were cortically scaled with eccentricity to stimulate approximatelyequal areas of visual cortex (ie, central sectors were smaller than peripheralsectors). The entire display subtended a diameter of 44.5°, and the central12 sectors fell within 2.6° of the foveal center (Figure 1A). The stimulus array was displayed on a black-and-whitemonitor driven at a frame rate of 75 Hz. The 16-element checkerboard of eachsector had a probability of 0.5 of reversing on any pair of frame changes,and the pattern of reversals for each sector followed a pseudorandom m−sequence.Three channels of recording were obtained using gold cup electrodes. The electrodesfor the midline channel were placed 4 cm above the inion (active), at theinion (reference), and on the forehead (ground). The same ground and referencewere used for the other 2 channels, but the active electrode was placed 1cm up and 4 cm lateral to the inion on either side. By subtracting differentcombinations of pairs of channels, 3 additional derived channels were obtained,resulting in effectively 6 channels of recording representing the 6 possiblepairs of the 4 recording electrodes.12,13
The VEP records were amplified with the high- and low-frequency cutoffsset at 3 and 100 Hz (preamplifier P511J; Grass Instrument Co, Quincy, Mass),and were sampled at 1200 Hz (every 0.83 milliseconds). The m−sequencehad 215 − 1 elements, requiring about 7 minutes of recording.Two 7-minute recordings were obtained during monocular stimulation of eacheye. Both eyes were tested twice in ABBA fashion, and the average of 2 recordingswas used for analysis. The second-order kernel responses were extracted usingcomputer software (VERIS 4.3; Electro-Diagnostic Imaging). The mfVEPs werelow-pass filtered using a sharp cutoff at 35 Hz and a fast Fourier transformtechnique. This and all other analyses were performed with programs writtenin computer software (MATLAB; MathWorks Inc, Natick, Mass). A refractor/camera(Electro-Diagnostic Imaging) was used to refract the subjects' eyes and monitoreye position and fixation stability. Subjects were asked to fixate on thecenter of a black "X" located in the center of the display. Segments contaminatedby eye movements, loss of fixation, and/or noise were discarded and rerecorded.
Root-mean-square (RMS) amplitudes were calculated for each mfVEP responseduring an interval from 45 to 150 milliseconds. The signal-noise ratio (SNR)was also measured for each response, as previously described.12- 14 Briefly,to obtain the SNR for an individual response, the RMS of the response from45 to 150 milliseconds was divided by an estimate of the noise in the records.This noise measure was obtained, for each eye of each individual, as the meanof the 60 RMS amplitudes of the records from 325 to 430 milliseconds, a regionof the record virtually without a signal (Figure 1B). The SNR is equal to the following: RMS(signal window)/meanRMS(noise window).
Two analyses were performed on the best of the responses from the 6channels, a monocular and an interocular analysis. The best responses foran individual were derived differently for the 2 analyses. For the monocularanalysis, for each eye and each location, the response with the largest SNR,among the 6 channels, was selected. The 60 responses thus chosen for eacheye defined the best array for that eye. For the interocular analysis, ateach location, the response with the largest SNR was selected among the 12responses (2 eyes × 6 channels). The response from the other eye fromthat same channel made up the pair of responses at that location in the interocularbest array. Locations where the larger of the monocular responses had an SNRbelow 1.7 were excluded from the analysis. This represented 1% of the responsesfrom the healthy controls.13
For the monocular test, to determine if the mfVEP responses were significantlysmaller, the SNR of an individual response was compared with the mean andstandard deviation of the SNRs for the 30 control subjects at that location.Probability plots, resembling the probability plots for the HVF test, wereproduced by coding whether the SNRs of the responses were significantly differentfrom those of healthy controls.12,15 Examplesof monocular probability plots for the left and right eyes of a patient areshown in Figure 2A and B, respectively.Each square locates the center of 1 of the 60 sectors of the stimulus display(Figure 1A). The colored squaresindicate the locations with SNR values that fell more than 1.96 (light color)or 2.58 (dark color) SDs below the mean values. Blue indicates that the righteye, and red that the left eye, had significantly smaller SNRs.
For the interocular analysis to determine if the mfVEP response wassignificantly smaller in one eye compared with the other, the ratio of theRMS amplitudes from the 2 eyes was calculated for each location (ie, log10 [RMS of the right eye/RMS of the left eye]).7,12 Thelog of the interocular ratio obtained from the patient for each location wasthen compared with the mean and standard deviation of the log ratio valuesfrom the control subjects, and an interocular probability plot was derived.An example of an interocular probability plot and of the mfVEP responses obtainedfrom the right (blue) and left (red) eyes of another patient is shown in Figure 2C and D, respectively. Each squarein the probability plot locates the center of 1 of the 60 sectors of the stimulusdisplay. The colored squares indicate the locations with values that fellmore than 1.96 (light color) or 2.58 (dark color) SDs below the mean values.For this patient, the right eye had significantly smaller signals than theleft eye (the squares are blue), and most of the responses in the inferiorhemifield were significantly decreased.
The optic nerve head was assessed with a confocal scanning laser ophthalmoscope(Heidelberg Retina Tomograph [HRT] II; Heidelberg Engineering GmbH, Heidelberg,Germany). The size of the field of view was 15° × 15°, and digitizationwas performed in frames of 384 × 384 pixels. The spatial resolutionwas 10 µm per pixel. Three 3-dimensional images were obtained for eacheye from each patient. With the HRT II, there is an automatic online qualitycontrol during image acquisition. If 1 or more of the acquired image seriescannot be used for any reason (eg, if the patient were to lose fixation),then additional images are automatically acquired until 3 useful image serieshave been obtained. The mean topographic result of the 3 scan series was usedfor analysis. The contour line was drawn around the optic disc by one of us(P.T.). The standard reference plane was used. There are several approachesthat can be used to discriminate between healthy and glaucomatous optic discs.These approaches rely on global and/or sectorial topographic indices and havesimilar sensitivities.16 We used the Moorfieldsregression analysis approach. It has the advantage of adjusting the globaland sectorial rim areas for disc size and age to improve specificity and toallow for the assessment of regional damage. Regional damage was assessedwith HRT II software by dividing the optic nerve head into 6 sectors (nasal,supranasal, supratemporal, temporal, inferotemporal, and inferonasal). Thesectors were compared with those in a normal database using Moorfields regressionanalysis, and then classified into 1 of 3 categories (within normal limits,borderline, or outside normal limits).17 Briefly,this classification was performed as follows: the rim and disc area for eachsector were compared with those in a normal database, and the sectors werethen classified depending on the patient's age and the overall size of theoptic disc. The analysis provided a predicted value and an actual value forthe rim area of each of the 6 sectors. If the percentage of the rim area fora given sector was larger than or equal to the 95% age-dependent confidencelimit, then the sector was classified as being within normal limits. If thepercentage of rim area was between the 95% and 99.9% confidence limits, thenthe respective sector was classified as borderline. Finally, if the percentageof the rim area for a sector was lower than the 99.9% confidence limit, thenthe sector was classified as being outside normal limits. Only those sectorsclassified as being outside normal limits were included in our mapping analysis.
The rim area sectors were related to HVF 24-2 locations using a mapbased on one developed by Garway-Heath et al.18,19 Themap relates the 6 sectors of the optic nerve head to the visual field testpoints (Figure 3A and B). The 6sectors were also related to the 60 mfVEP probability plot locations (Figure 3C).
The HVF probability plots (total deviation plots), HRT rim sector maps,and monocular and interocular mfVEP probability plots were compared for eachpatient. An example of the format developed for displaying the results isshown in Figure 4. The isodegreecontours (radii of 2.6° and 22.2°) have been drawn to make it easierto compare the HVF probability plots (Figure4A and B), the HRT rim sector maps (Figure 4C and D), and the mfVEP probability plots (Figure 4E-G). For this patient, the HVF probability plots show anupper hemifield defect in the left eye and no significant defects in the righteye (Figure 4A and B, respectively).The saturated colored squares indicate those locations with P<.01 vs the age-matched controls, and the desaturated colored squaresindicate those locations with P<.05 vs the controls.
To provide a similar topographic map for the HRT results, the map in Figure 3A was coded as shown in Figure 4C and D. If a sector was classifiedas being outside the normal limits, then all the corresponding points in theHVF were circled, as in Figure 4C.A colored circle indicates that the percentage of the rim area for the correspondingsector was lower than the 99.9% confidence limit. In this case, the HRT mapspredict a significant defect in the upper hemifield of the left eye and nosignificant defects in the right eye. The monocular (Figure 4E and F) and interocular (Figure 4G) mfVEP probability plots also show an upper hemifielddefect in the left eye and no significant defects in the right eye. For thispatient, all 3 tests are in agreement. There are slight differences in theextent of the defect between the 3 measures that can be attributed to differencesin the way the visual field is sampled. For example, on the mfVEP plot, thedefect extends into the central 2.6°, but some of the outermost sectorsappear to be normal. With the mfVEP technique, 12 responses are obtained inthe central 2.6°, compared with 1 measure of sensitivity for the foveawith the HVF.
Figure 5 shows the resultsobtained from a patient who showed agreement between the 2 measures of visualfunction. The HVF (Figure 5A andB) and mfVEP (Figure 5E-G) resultsshow significant defects in the right eye (Figure 5B, F, and G), while the HRT results (Figure 5C and D) were classified as being within the normal limits.A lower hemifield defect can be seen in the HVF probability plot for the righteye (Figure 5B), together with 2points in the upper field with P<.01. There areno significant defects in the plot for the left eye (Figure 5A). The HRT maps (Figure5C and D) predict no significant defects in either eye. In agreementwith the HVF, the mfVEP shows a lower hemifield defect for the right eye (Figure 5F), and there is also a cluster of3 points at P<.01 in the upper hemifield on theinterocular probability plot (Figure 5G).
The results of the 3 tests for all 20 patients (40 eyes) are summarizedin Figure 6 and in Table 1 and Table 2.In Figure 6 and in Table 1 and Table 2,the locations of the HVF and mfVEP defects are related to the 6 rim area sectorsshown in Figure 3B. Sectors on theHVF and mfVEP probability plots were classified as being outside the normallimits if they contained either 2 or more adjacent field locations with P<.01 or 3 or more adjacent locations with P<.05. Of a total of 240 sectors tested (40 eyes × 6 sectors),45 (18.8%) on HVF, 53 (22.1%) on mfVEP, and 26 (10.8%) on HRT testing weresignificantly different from those of control subjects. Twice as many sectorswere classified as abnormal with the 2 measures of visual function than withthe measure of structural damage.
A comparison between these 2 measures of visual function (HVF and mfVEP)shows that there is agreement for 86.7% of the sectors: 175 (72.9%) of thesectors had no significant defects and 33 (13.8%) had abnormalities (Table 1). Of the disagreements betweenthe tests, 20 (8.3%) of the sectors showed significant mfVEP defects and noHVF defects; 13 of these sectors corresponded to the central visual fieldarea, or sector 1. To compare these functional measures with the HRT, theHVF, and the mfVEP, results were separated into 3 groups according to whether(1) there was agreement about the presence of a defect in a sector, (2) therewas agreement about the absence of a defect (the sector was normal), or (3)there was disagreement between the measures. The correspondence between these3 groups and the HRT results is shown in Table 2. Of the 208 sectors showing agreement on the functionalmeasures, the HRT results were in agreement in 176 (84.6%) of the cases; 165sectors had no significant defects on all 3 measures, while 11 had significantdefects. Thus, in 32 (15.4%) of the cases in which the mfVEP and HVF resultsagreed, the HRT showed discrepant results. This comparison is potentiallymisleading because only 26 (10.8%) of the 240 sectors had significant defectson HRT testing compared with about 20% for the HVF and the mfVEP. In our HRTmapping analysis, we only included sectors classified as being outside thenormal limits. If we were to change our criterion and include sectors classifiedas borderline, then 76 (31.7%) of the sectors would have defects on HRT testing.However, this would not improve the agreement between the functional and structuraltests—it would decrease to 74% (Table3).
The development of new technologies designed to measure the optic nerveand nerve fiber layer and new techniques such as the mfVEP to assess visualfunction promise to increase our understanding of the relationship betweenstructural and functional abnormalities in glaucoma. However, to realize thispromise, we need to develop methods for comparing the results of these newtechniques. In addition, a topographic comparison of the results will enhancethe clinician's ability to diagnose and monitor disease progression. The methodpreviously described allows for such a comparison between 2 measures of visualfunction, the HVF and the mfVEP, and a measure of the optic nerve head integrity,the HRT. Specific HVF and mfVEP locations suggesting glaucomatous damage arerelated to structural changes in the neural rim of the optic nerve head assessedwith HRT.
Previous studies20- 22 haveprovided methods for mapping structural damage to functional damage assessedwith achromatic automated perimetry or with short-wavelength automated perimetry.A review is provided by Johnson et al.23 Thesemaps showed that some visual field zones topographically mapped to certainrim sectors (eg, patients with superior hemifield sensitivity loss tendedto have inferior rim defects, and vice versa). In addition, one defectivevisual field zone may be related to several rim sectors.
In this study, we used the map developed by Garway-Heath et al18 to relate the test locations of the HVF and the mfVEPto sectors of the optic nerve head measured with the HRT II. Despite the factthat there are differences between the 2 functional tests regarding stimulusduration, stimulus size, and adaptation level, and more important differencesin the mechanism of response generation and in the nature of the responsemeasure, the agreement between the 2 measures of visual function was quitegood. This finding is consistent with previous work.6,12 Thedisagreements between the 2 measures are interesting. Most sectors definedas abnormal on the mfVEP but normal on the HVF were located in the centerof the visual field; they corresponded to sector 1. One explanation for thisfinding is that the field is sampled in a different way by the 2 techniques.In particular, within the central 2.6° (5.2° diameter), there are12 mfVEP responses but only 1 HVF test point. This ability of the mfVEP techniqueto sample the central retina in detail may be relevant to findings implicatingmacular loss in glaucoma. Whether the technique is more sensitive in detectingdefects in the central region of the visual field than the HVF 10-2, however,remains to be tested.
Despite reasonably good overall agreement between the HRT and the 2measures of visual function, there was disagreement for a number of cases.If we consider the 208 sectors that showed agreement between the HVF and themfVEP, we find discrepancies with the HRT results for 32 (15.4%) of thesesectors. If we include the borderline HRT results as abnormal, this only leadsto a decrease in the number of sectors showing agreement. Figure 5 shows one example in which the functional tests show aclear defect that is missed on the HRT. Thus, there are clearly cases in whichthe functional measures and the structural measure disagree.
In summary, a method for comparing mfVEP responses with behaviorallydetermined visual fields and with structural measures of the optic nerve headhas been developed. The preliminary results suggest that, while in generalthe structural and functional measures agree, there are clear discrepanciesthat merit further study.
Correspondence: Vivienne C. Greenstein, PhD, Department of Ophthalmology,New York University School of Medicine, 550 First Ave, New York, NY 10016(email@example.com).
Submitted for publication March 28, 2003; final revision received March4, 2004; accepted March 4, 2004.
This study was supported in part by grant EY02115 from the NationalEye Institute, National Institutes of Health, Bethesda, Md; an unrestrictedgrant from Research to Prevent Blindness Inc, New York, NY (Departments ofOphthalmology, Colombia University and New York University School of Medicine);and the Steven and Shelley Einhorn Research Fund of the New York GlaucomaResearch Institute, New York.
This study was presented in part at the Annual Meeting of the Associationfor Research in Vision and Ophthalmology; May 4, 2003; Fort Lauderdale, Fla.