Angioscotoma detection with fundus-oriented perimetry A ... · horizontal black and gray lines show...

13
Vision Research 39 (1999) 1897 – 1909 Angioscotoma detection with fundus-oriented perimetry A study with dark and bright stimuli of different sizes U. Schiefer a, *, N. Benda b , T.J. Dietrich a , B. Selig a , C. Hofmann c , J. Schiller a a Uni6ersity Eye Hospital, Dept. II, Schleichstraße 12 -16, D-72076 Tu ¨bingen, Germany b Department of Medical Biometry, Westbahnhofstraße 55, D-72070 Tu ¨bingen, Germany c Institute of Applied Physics, Auf der Morgenstelle 10, D-72076 Tu ¨bingen, Germany Received 21 November 1997; received in revised form 2 October 1998 Abstract Fundus-oriented perimetry (FOP) was used to evaluate the effectiveness of different-sized bright and dark stimuli in detecting and quantitatively measuring angioscotoma. The foveolas and optic disks of digitized fundus images were aligned with their psychophysical counterparts to construct individual grids of perimetric stimuli. Each grid included a linear set of test point locations crossing a retinal vessel. Angioscotomas immediately became visible in nine of 13 healthy normal volunteers tested with FOP. Additional mathematical processing of local loss of differential light sensitivity (dls) disclosed an angioscotoma for at least one stimulus condition in all persons tested. The angioscomas were usually deeper for small (12%) targets than for large (32%) ones. On the other hand, the overall noise at dls thresholds was generally higher for small than for large stimuli regardless of whether the stimuli were bright or dark. No noteworthy differences were found in detection rates or signal-to-noise ratios under different stimulus conditions (dark/bright/small/large). FOP permits the individual arrangement of stimuli for specific morphological conditions and is thus capable of detecting even minute visual field defects such as angioscotomas. © 1999 Elsevier Science Ltd. All rights reserved. Keywords: Angioscotoma; Fundus-oriented perimetry; Stimulus size; Dark stimulus; Signal-to-noise ratio 1. Introduction Angioscotomas are usually regarded as a nuisance, particularly because they resemble unstable, minor nerve fiber bundle defects. Because of their small size, however, angioscotomas can serve as accurate indica- tors of the quality of perimetric procedures, especially regarding spatial resolution. This is important in the functional assessment and follow-up of minute mor- phological retinal lesions, e.g. slit-like nerve fiber layer defects or circumscribed vascular or inflammatory pro- cesses. Recent publications and our own unpublished results indicate that regional concentrations of stimuli are useful and effective in detecting circumscribed mor- phological lesions (Tuulonen, Lehtola & Airaksinen, 1993; Be ´chetoille, 1996; Langerhorst, Carenini, Bakker & De Bie-Raakman, 1997; Orzalesi, Miglior, Lonati & Rossetti, 1998). Fundus-oriented perimetry (FOP) can correlate morphology with function by individually ‘tailoring’ perimetric test point arrangements to ophthalmo- scopic fundus findings. In addition, stimulus density can be increased locally, and numerous reversals are possible. This ensures precision and reliability in the psychophysical measurement of circumscribed re- gions of interest. FOP can be carried out with both bowl perimeters and flat screens, i.e. in perimetry and campimetry. However, in contrast to con- ventional perimetric methods, in which test points must be brighter than the surrounding background, FOP monitors can present dark (‘off’) stimuli as well. This study addressed the following questions: (1) Does target size affect the depth and width of angioscotomas? (2) Which type of stimulus yields the best signal- to-noise ratio for angioscotomas, i.e. the best ratio of scotoma depth to threshold variability? * Corresponding author. Fax: +49-7071-295038; e-mail: ul- [email protected]. 0042-6989/99/$ - see front matter © 1999 Elsevier Science Ltd. All rights reserved. PII:S0042-6989(98)00295-8

Transcript of Angioscotoma detection with fundus-oriented perimetry A ... · horizontal black and gray lines show...

Page 1: Angioscotoma detection with fundus-oriented perimetry A ... · horizontal black and gray lines show the methodologically achievable minimum of differential light sensitivity for black

Vision Research 39 (1999) 1897–1909

Angioscotoma detection with fundus-oriented perimetryA study with dark and bright stimuli of different sizes

U. Schiefer a,*, N. Benda b, T.J. Dietrich a, B. Selig a, C. Hofmann c, J. Schiller a

a Uni6ersity Eye Hospital, Dept. II, Schleichstraße 12-16, D-72076 Tubingen, Germanyb Department of Medical Biometry, Westbahnhofstraße 55, D-72070 Tubingen, Germany

c Institute of Applied Physics, Auf der Morgenstelle 10, D-72076 Tubingen, Germany

Received 21 November 1997; received in revised form 2 October 1998

Abstract

Fundus-oriented perimetry (FOP) was used to evaluate the effectiveness of different-sized bright and dark stimuli in detectingand quantitatively measuring angioscotoma. The foveolas and optic disks of digitized fundus images were aligned with theirpsychophysical counterparts to construct individual grids of perimetric stimuli. Each grid included a linear set of test pointlocations crossing a retinal vessel. Angioscotomas immediately became visible in nine of 13 healthy normal volunteers tested withFOP. Additional mathematical processing of local loss of differential light sensitivity (dls) disclosed an angioscotoma for at leastone stimulus condition in all persons tested. The angioscomas were usually deeper for small (12%) targets than for large (32%) ones.On the other hand, the overall noise at dls thresholds was generally higher for small than for large stimuli regardless of whetherthe stimuli were bright or dark. No noteworthy differences were found in detection rates or signal-to-noise ratios under differentstimulus conditions (dark/bright/small/large). FOP permits the individual arrangement of stimuli for specific morphologicalconditions and is thus capable of detecting even minute visual field defects such as angioscotomas. © 1999 Elsevier Science Ltd.All rights reserved.

Keywords: Angioscotoma; Fundus-oriented perimetry; Stimulus size; Dark stimulus; Signal-to-noise ratio

1. Introduction

Angioscotomas are usually regarded as a nuisance,particularly because they resemble unstable, minornerve fiber bundle defects. Because of their small size,however, angioscotomas can serve as accurate indica-tors of the quality of perimetric procedures, especiallyregarding spatial resolution. This is important in thefunctional assessment and follow-up of minute mor-phological retinal lesions, e.g. slit-like nerve fiber layerdefects or circumscribed vascular or inflammatory pro-cesses. Recent publications and our own unpublishedresults indicate that regional concentrations of stimuliare useful and effective in detecting circumscribed mor-phological lesions (Tuulonen, Lehtola & Airaksinen,1993; Bechetoille, 1996; Langerhorst, Carenini, Bakker& De Bie-Raakman, 1997; Orzalesi, Miglior, Lonati &Rossetti, 1998).

Fundus-oriented perimetry (FOP) can correlatemorphology with function by individually ‘tailoring’perimetric test point arrangements to ophthalmo-scopic fundus findings. In addition, stimulus densitycan be increased locally, and numerous reversals arepossible. This ensures precision and reliability inthe psychophysical measurement of circumscribed re-gions of interest. FOP can be carried out with bothbowl perimeters and flat screens, i.e. in perimetryand campimetry. However, in contrast to con-ventional perimetric methods, in which test pointsmust be brighter than the surrounding background,FOP monitors can present dark (‘off’) stimuli aswell.

This study addressed the following questions:(1) Does target size affect the depth and width of

angioscotomas?(2) Which type of stimulus yields the best signal-

to-noise ratio for angioscotomas, i.e. the best ratioof scotoma depth to threshold variability?

* Corresponding author. Fax: +49-7071-295038; e-mail: [email protected].

0042-6989/99/$ - see front matter © 1999 Elsevier Science Ltd. All rights reserved.

PII: S 0 0 4 2 -6989 (98 )00295 -8

Page 2: Angioscotoma detection with fundus-oriented perimetry A ... · horizontal black and gray lines show the methodologically achievable minimum of differential light sensitivity for black

U. Schiefer et al. / Vision Research 39 (1999) 1897–19091898

(3) To what extent do angioscotoma differ from the‘underlying’ morphological cause?

2. Methods

FOP superimposes an individually designed perimet-ric grid onto the digitized image of an ocular fundus.The procedure has been described in detail elsewhere(Schiefer & Witte, 1996; Schiefer, Stercken-Sorrenti,Dietrich, Friedrich & Benda, 1996a; Schiefer, Stercken-Sorrenti, Dietrich, Selig, Benda & Friedrich, 1996b). Itis basically adaptable for all conventional bowl-typeprojection perimeters, provided that they are capable ofimporting digital image data from a network, diskette,photo-CD, or other storage medium.

The procedure has been incorporated into the Tubin-gen Computer Campimeter (TCC). It was run in thisstudy on a Power PC (Apple Inc., Cupertino, CA) towhich three monitors were connected:

(1) The ‘Stimulus monitor’: Stimuli were presentedon a high-resolution, true-colour monitor (CALIBRA-TOR, Barco Inc., Kortriyk, Belgium). The monitor hadbeen calibrated by means of a mobile luminance meterto ensure homogeneity of the background and lumi-nance stimuli over the entire surface of the monitor(Dietrich et al. (1996). Maximum luminance of themonitor after calibration was 68 cd m−2, and minimumluminance was 0.2 cd m−2. A four-point (‘red dia-mond’) fixation target was gradually faded into thecenter of the screen. Stimulus duration was 200 ms.

(2) The fixation monitor: fixation was supervised withthis video display unit. An infrared (IR) video cameraattached to the chin and head rest permitted depictionof gaze movements on the monitor. In addition, thesignals of the video camera were sampled every 40 msby a specially designed computer graphic board whichrecorded position and horizontal diameter of the pupil.There was no automatic correction for eye movements.

(3) The control monitor: this monitor showed actua-tion of the examination process, displayed the perimet-ric grid, and showed the progress and results of theexamination. The monitor was also used to depict andprocess digitized fundus images and to create individualstimulus grids by superimposing them.

As mentioned, FOP correlates morphology and func-tion by using two characteristic landmarks (i.e. theblind spot and the center of the visual field, Bek &Lund-Andersen, 1990). In this way a digitized fundusimage can be graphically displayed and processed bymeans of inversion along the horizontal and verticalaxes. After superimposition, the center of the visualfield is manually aligned horizontally or vertically withthe foveola of the digitized fundus image. The blindspot, which has been initially identified by manualperimetry, is then superimposed onto the optic disk of

the digitized fundus image. This activates rotatory andzoom functions for the fundus image. Subsequently, afundus-oriented arrangement of test points can be indi-vidually created on the screen of the control monitor.

In this study, 24 test points 6% apart were arranged ina line across a single retinal vessel on digitized fundusimages of young normal subjects. The profile sectionwas nearly perpendicular to the center of the visualfield, making it possible to describe the position of thevessel as a function of eccentricity. Another 24 testpoints were positioned along the vertical or horizontalmeridians, and two more stimuli were situated in thecenter and near the blind spot region of the visual field(see Fig. 1, bottom). The size of the stimuli was set to12% or 32%. Both bright (increment) and dark (decre-ment) targets were presented, resulting in four stimulusconditions: 12%-bright, 12%-dark, 32%-bright, and 32%-darktest points.

The superimposed fundus image also provided alandmark for manual perimetry or ‘angioscotometry’ ofthe vascular region with a very small (8%), bright target(see arrows in Fig. 1 and Fig. 2a and b).

The examination itself began after digital storage ofthe individually created perimetric grid. A total of 12%of the stimulus presentations (i.e. 4% each for false–positive, false–negative, and fixation checks) were usedas catch trials. Fixation was checked by presentingtargets 5 dB above the initially estimated dls thresholdat the center of the visual field. A maximum of twoperimetric examinations was carried out during eachsession. Examination sequences were randomized usingthe four stimulus conditions mentioned above.

Differential light sensitivity (dls) levels vary in theliterature, since baseline values differ from one publica-tion to another. However, this shift is only linear, sincea logarithmic scale is used, and the relative depth ofangioscotomas in relation to the surrounding dls leveltherefore remains comparable for both dark and brightstimuli.

Differential light sensitivity was calculated in thepresent study with a modified Weber’s formula(Menage, 1997) by relating the absolute luminance dif-ference between the actual stimulus and the back-ground (�DL �) to a the background luminance of 10 cdm−2. The absolute value of DL made it possible todefine dls results for both bright and dark stimuli withone and the same formula:

dls [dB]=10× log (10 cd m−2/ �DL �).A baseline value of 10 cd m−2 was chosen to yield

identical dB-values for dark and bright test points withidentical absolute luminance differences in relation tobackground luminance.

The dB scale for dark stimuli is physically limited to0 dB (minimum luminance of the monitor for ‘blacktest points’=maximum contrast). According to the

Page 3: Angioscotoma detection with fundus-oriented perimetry A ... · horizontal black and gray lines show the methodologically achievable minimum of differential light sensitivity for black

U. Schiefer et al. / Vision Research 39 (1999) 1897–1909 1899

Fig. 1. Bottom: fundus image of a normal subject with superimposed perimetric grid. Top: Profile of differential light sensitivity (dls) along a lineof test points crossing a retinal vessel. The angioscotoma, i.e. the dip in the dls profile, is marked by a + . The arrow shows the scotoma positiondetermined by manual kinetic perimetry, the circle and crosshairs show the location and diameter of the retinal vessel derived topographically fromthe fundus image with the Littmann formula (Littmann, 1988), see also Fig. 2a and b.

above-mentioned formula, bright stimuli with a lumi-nance difference of more than 10 cd m−2 are character-ized by negative dB values. Differential light sensitivity

never exceeded −7.00 dB in the present study becauseof the restricted luminance level of the TCC monitor(560 cd m−2). Local dls was designated by * and was

Page 4: Angioscotoma detection with fundus-oriented perimetry A ... · horizontal black and gray lines show the methodologically achievable minimum of differential light sensitivity for black

U. Schiefer et al. / Vision Research 39 (1999) 1897–19091900

Fig. 2. Profile sections of dls along lines of test points in the regions of the retinal vessels in another 12 normal subjects under various stimulusconditions. Obvious angioscotomas for at least one profile section are marked by + , those with a questionable depression of local dls by ?. Thehorizontal black and gray lines show the methodologically achievable minimum of differential light sensitivity for black and white stimuli,respectively. Local dls was marked by * and set to −10 dB for all situations, in which none of the bright stimulus presentations at a givenlocation were perceived. Local dls was marked by * and set to −7 dB for all situations, in which none of the dark stimulus presentations at agiven location were perceived.

Page 5: Angioscotoma detection with fundus-oriented perimetry A ... · horizontal black and gray lines show the methodologically achievable minimum of differential light sensitivity for black

U. Schiefer et al. / Vision Research 39 (1999) 1897–1909 1901

Fig. 2. (Continued)

Page 6: Angioscotoma detection with fundus-oriented perimetry A ... · horizontal black and gray lines show the methodologically achievable minimum of differential light sensitivity for black

U. Schiefer et al. / Vision Research 39 (1999) 1897–19091902

Table 1Mean, median and maximal proportions of false answers to false positive, false negative and fixation controls for the four given stimulusconditions.

12% bright 32% dark12% dark 32% bright

0.02 0.03Mean 0.010.020.00 0.000.000.00False-positive Median

0.08 0.11Maximum 0.12 0.06

0.020.02 0.000.04Mean0.00 0.00False-negative Median 0.00 0.00

0.000.140.200.20Maximum

0.12 0.01Mean 0.03 0.010.000.000.050.00False control Median

0.63 0.05Maximum 0.050.29

set to −10 dB whenever none of the bright stimuliwere perceived at a given location (see left ordinate); itwas designated with * and set to −7 dB when no darkstimuli were perceived at a given location (see rightordinate). However, values for deeper visual field de-fects would have been possible for both types ofstimuli.

A modified 4-2-1 strategy with four reversals (the‘two- button yes-/no- method of inquiry’, see Lutz,Dietrich, Benda, Selig, Schiefer & Daum, 1996) wasused in this study. Thresholds were assessed with themaximum likelihood method, based on a logistic regres-sion model (‘logit analysis’). Two procedures were usedalternately:

(a) Single threshold estimation: Each dls thresholdwas calculated at every stimulus location, and the indi-vidual results were then collated (see Fig. 1, top, andFig. 2a and b). Instances of circumscribed dls depres-sion (\5 dB) in at least one profile section were ratedas obvious angioscotoma and marked with + . Thevalue of 5 dB was chosen since as it also serves as acut-off criterion in conventional threshold-oriented,supraliminal perimetry. Instances of a localized depres-sion between 2.5 and 5 dB in at least one profile sectionwere rated as questionable angioscotomas and markedwith a question mark (?).

(b) One-step angioscotoma estimation: The probabil-ity of test stimulus perception was described as a func-tion of intensity and location (see Figs. 4–8), forexample as follows:

If 8 is the distance from a test point on a line acrossa retinal vessel to a fixed point of reference on this line,the threshold near the vessel can be described by:

m(8)=a−b 8+h�

tanh��8−c

d/2�2n

−1�

where a−b 8 describes the linear trend of thethreshold near the vessel and next to the scotoma, c thecenter of angioscotoma, and h the scotoma depth at thecenter. Scotoma width, defined here as the diameter of

the local depression at 24% of its maximal depth(Benda, Schiefer & Dietrich, 1996), is given by d.Stimulus detection can be described by a logistic regres-sion model with threshold m dependent on stimuluslocalization as described above. s, representing theglobal spread of the logistic regression, provides ameasurement of threshold variability or the patient’suncertainty in describing the ‘noise’ during the an-gioscotoma search. Thus the relation of scotoma depthh to global spread s provides a measure of signal-to-noise ratio (i.e. h/s), cf. Benda et al., 1996.

Statistical analysis was carried out with the programJMP™ 3.1.6 program (SAS Institute Inc., Cary, NC,USA).

3. Results

3.1. Single thresholds estimations

Angioscotomas were clearly visible for at least onestimulus condition in nine of the 13 subjects and weretherefore rated as detected by FOP (Fig. 1 and Fig. 2aand b). Each of the remaining four cases showed ques-tionable angioscotomas for at least one stimulus condi-tion. A misalignment in test point arrangementoccurred in one instance (Patient 3) due to a dis-crepancy between the results of FOP and manualperimetry (angioscotometry).

Individual dls estimations did not differ notably forbright and dark stimuli of a given diameter (see Fig. 1and Fig. 2). As the figures show, dls values weregenerally higher for 32%-stimuli than for 12%-stimuli.This was true for both bright and dark test points.Especially in the case of the 12%-stimuli it must beremembered that the dynamic range was methodologi-cally restricted (dark stimuli: minimum value 0 dB,bright stimuli: minimum value :−7.00 dB!).

Table 1 shows mean, median and maximum proper-ties of false answers to false positive, false negative and

Page 7: Angioscotoma detection with fundus-oriented perimetry A ... · horizontal black and gray lines show the methodologically achievable minimum of differential light sensitivity for black

U. Schiefer et al. / Vision Research 39 (1999) 1897–1909 1903

Fig. 3. Retest reliability of FOP: Five consecutive examinations (dark stimuli, 12%) of one person (subject 8); all were carried out on another day(see also Fig. 2).

fixation checks. The median seems to be most effectivein comparing these results but shows no relevantchanges for the four given stimulus conditions.

Using a 12% dark stimulus, FOP found the sameangioscotoma in each of five reliability tests conductedonce daily on one volunteer (Subject 8) for 5 successivedays. As Fig. 3 shows, scotoma positions varied by lessthan 1° in this subject.

3.2. One-step angioscotoma estimation

Since evaluation of profile sections is inevitably quitesubjective, an alternative method (Benda et al., 1996)

was applied: the entire curve of a profile section wasestimated on the basis of each single measurementvalue (see Figs. 4–8). This procedure was free of sub-jective interpretation. It proved capable of describing ascotoma under all four stimulus conditions in sevensubjects, and under at least one stimulus condition inall tested persons. It outlined a scotoma in 12 subjectswhen the bright 32%-stimulus was used and in ten sub-jects under all the other stimulus conditions (see Figs.4–6).

Angioscotoma depths decreased as stimulus diameterincreased. This was true of all bright and most dark testpoints (Fig. 4). Friedman’s two-way ANOVA showed a

Page 8: Angioscotoma detection with fundus-oriented perimetry A ... · horizontal black and gray lines show the methodologically achievable minimum of differential light sensitivity for black

U. Schiefer et al. / Vision Research 39 (1999) 1897–19091904

Fig. 4. Angioscotoma depth estimations (h) of 13 normal subjects with bright and dark stimuli of various diameters in the ‘one-step estimation’procedure (see text). Note that angioscotoma description was impossible under some conditions.

significant difference for the four stimulus conditions(n=0.034). The relative approximate standard error ofmaximum likelihood estimation ranged from 7 to 75%of the corresponding estimate (median: 19%).

When bright stimuli were used, angioscotoma widthgenerally increased as stimulus diameters increased.However, there was no uniform trend in the case ofdark targets (Fig. 5). Angioscotoma width was usuallysomewhat greater than the diameter of the causativeretinal vessel, which was estimated with the Littmannformula (Littmann, 1988). This was true in eight of 11cases with the bright 32%-stimuli and in nine of tenevaluable cases with the dark 12%-stimuli. However,Friedman’s two-way ANOVA did not show significantdifferences among the four stimulus conditions (n=0.068).

With few exceptions, the calculated positions of ves-sels corresponded well with subjectively estimated loca-tions of angioscotomas (Fig. 6; see also Fig. 1 and Fig.2a and b). When estimated by kinetic perimetry andFOP, the deviation was never more than 1.3° and didnot exceed 0.5° in most cases (eight of 13; Fig. 6).

The noise of profile sections was estimated by theglobal spread parameter s via logistic regression (seeFig. 7). There were significant differences among thefour stimulus conditions (n=0.00046; Friedman’s two-way ANOVA). In most cases, noise was lower for largestimuli than for small ones; this was true of both brightand dark targets. The relative approximate standarderror of maximum likelihood estimation ranged from13 to 20% (median 15%) of the corresponding estimate.

However, signal-to-noise ratio (h/s) showed no con-sistent dependency on target size (32% vs. 12%) and targetluminance characteristics (i.e. bright vs. dark) (Fig. 8).Friedman’s two-way ANOVA yielded a n-value of 0.69for the four stimulus conditions. The relative approxi-mate standard error of maximum likelihood estimationranged from 15 to 77% (median 23%) of the corre-sponding estimate.

4. Discussion

Perimetric assessment of angioscotomas was onceregarded as a clinically important diagnostic procedurefor acquiring additional information on intraocularpressure (glaucoma), intracranial pressure, and vascularor hemo-rheological conditions (Dashevsky, 1938;Welt, 1945; Goldmann, 1947; Abe, 1968). At present,however, visual field defects caused by retinal vesselstend to be viewed as physiologically induced nuisanceeffects (Chauhan, Henson & Hobley, 1988; Safran,Halfon, Safran & Mermoud, 1995). Nonetheless, thesmall size of angioscotomas makes them useful forassessing the quality of perimetry with high spatialresolution (Haberlin, Funkhouser & Fankhauser, 1983;Zulauf, 1988, 1990a,b; Fitzke & McNaught, 1994;Safran et al., 1995; Remky, Beausencourt & Elsner,1996a). This is important for functional assessment andthe follow-up of minute morphological retinal lesions,e.g. slit-like nerve fiber layer defects or circumscribedprocesses of a vascular or inflammatory nature. How-ever, it should be noted that there are differences in

Page 9: Angioscotoma detection with fundus-oriented perimetry A ... · horizontal black and gray lines show the methodologically achievable minimum of differential light sensitivity for black

U. Schiefer et al. / Vision Research 39 (1999) 1897–1909 1905

Fig. 5. Angioscotoma width estimations of 13 normal subjects with bright and dark stimuli of various diameters using the ‘one-step estimation’procedure. The diameters of retinal vessels (derived topographically from the fundus image with the Littman formula (1988)) are marked by �(see also Fig. 4).

‘pathogenesis’: the visual field defects just mentionedresult from a structural and/or functional impairmentof retinal elements, whereas a masking of (normal)photoreceptors takes place in angioscotomas.

Weekers and Humblet (1945) stated that ‘‘…a sco-toma—corresponding to the projection of a single reti-nal vessel in the field—is no longer physiological andhas to be regarded with certainty as a pathologicalneuro-scotoma as soon as it becomes as large as 2–3°…’’. The scotomas detected in the present studyfulfilled this requirement: their maximum width, asdefined by the formula used in one-step angioscotomaestimation, was 2° (see Fig. 5).

Perimetric efficiency increases when test points areconcentrated in circumscribed regions of interest. Thishas recently become feasible via FOP, which utilizesindividual morphological data (Schiefer & Witte, 1996;Schiefer et al., 1996a,b). However, this method may beaffected by artifacts from eye or head movements, sinceit is not directly related to the actual fundus image.This is all the more so in the presence of fixationinstability (Eizenman, Trope, Fortinsky & Murphy,1992; Rohrschneider, Becker, Kruse, Fendrich & Vol-cker, 1995a; Safran et al. 1995; Schiefer et al. 1996a).On the other hand, minimum (often rotatory) eyemovements (tremor: 10–15¦, drifts: 1–6% as well as‘microsaccades’ (Baumgartner, Bornschein, Hanitzsch,Jung, Kornhuber, Rentschler et al., 1978)) are essentialfor avoiding local adaptation and the Troxler phe-nomenon, respectively (Troxler, 1804; Cibis, 1965; Spill-mann & Kurtenbach, 1992).

Although campimetry can ensure exact presentationof test points on a monitor, precision in aiming at themwas restricted in this study by the comparatively smallscreen of the control monitor: a shift of only one pixelon this monitor resulted in a displacement of 0.2° (thesize of a minor retinal vessel) on the stimulus monitor.This threatened both the quality of kinetic perimetryand the superimposition of a perimetric grid onto agiven fundus image. Since no more than two examina-tions were carried out in any session, additional dis-placements were also possible due to maladjustment.Nevertheless, re-tests in a healthy subject on 5 differentdays confirmed that FOP has good reproducibility (seeFig. 3). The precision of the method was also under-scored by the fact that scotoma locations found withmanual (‘static’) perimetry and FOP usually differed byless than 1°.

Although perimetric procedures with direct fundusmonitoring—e.g. with the SLO—are capable in princi-ple of avoiding the above-mentioned problems (Tim-berlake, Mainster, Webb, Hughes & Trempe, 1982;Schneider, Kuck, Inhoffen & Kreissig, 1993; Sjaarda,Frank, Glaser, Thompson & Murphy, 1993; Sturmer,1993; Miglior, Rossetti, Brigatti, Bujtar & Orzalesi,1994; Hudson, Frambach & Lopez, 1995;Rohrschneider, Fendrich, Becker & Volcker, 1995b;Straub, Kroll & Kuchle, 1995; Sunness, Schuchard,Shen, Rubin, Dagnelie & Haselwood, 1995; Remky etal., 1996a; Remky, Elsner, Morandi & Beausencourt,1996b; Nishida & Kani, 1997; Rohrschneider, Gluck,Burk, Kruse & Volcker, 1997), they cannot be directly

Page 10: Angioscotoma detection with fundus-oriented perimetry A ... · horizontal black and gray lines show the methodologically achievable minimum of differential light sensitivity for black

U. Schiefer et al. / Vision Research 39 (1999) 1897–19091906

Fig. 6. Angioscotoma eccentricity in 13 normal subjects under various stimulus conditions using the ‘one-step estimation’ procedure. Scotomapositions determined by manual perimetry and the locations and diameters of retinal vessels (derived topographically from fundus image with theLittmann formula (Littmann, 1988)) are marked by � and �, respectively (see also Fig. 4).

adapted to conventional perimeters and continue tostruggle with technical restrictions (comparatively smallfield of examination, structurally inhomogeneous back-ground). Moreover, even SLO perimetry is affected byeye movements (Remky et al., 1996a). Automatic fun-dus tracking programs to correct such deviations ‘on-line’ before and during stimulus presentation are still inthe development stage.

The sizes and depths of angioscotomas induced bymajor retinal vessels in this study corresponded wellwith results obtained in other studies which used con-ventional white-on-white perimetry (bright test points)(Weekers & Humblet, 1945; Haberlin et al., 1983; Zu-lauf, 1988, 1990a,b; Safran et al., 1995). Scotomawidths derived ‘visually’ in this study from profile sec-tions (single threshold estimation, Fig. 1 and Fig. 2aand b) were somewhat smaller than those found byZulauf (1988) with 1–2°. This may have been due tothe higher resolution (‘mesh density’) of test pointarrangements used in the present study. Angioscotomawidths estimated by the above mentioned mathematical‘one step’ procedure in the present study tended to beeven smaller. However, the mathematical steps used todescribe angioscotoma width here were adjusted to apre-defined depth and therefore did not assess the totaldiameter of the visual field defects (see Section 2).

Haberlin et al. (1983) found angioscotoma widths onthe order of 0.6°, and angioscotoma depths of up to 10dB. Zulauf (1988) found angioscotoma depths of up to8 dB. Both in his report and in the present study,increasing the diameters of stimuli usually resulted in

minor local dls reductions in the angioscotoma region(see Fig. 1, Fig. 2 (a and b) and Fig. 4). This wasprobably due to the fact that test targets with diametersequal to or greater than those of retinal vessels under-going examination stimulate sensitive retinal regionsjust outside the deepest shadow, i.e. in the ‘penumbra’region of the vascular structure. A similar phenomenonhas been described in the case of glaucomatous visualfield defects (Zalta & Burchfield, 1990). Bek and Lund-Andersen (1989) showed comparable results for detec-tion of the blind spot with stimuli of different sizes.However, scotoma areas in the above-mentioned publi-cations were much larger than those caused by retinalvessels investigated in this study.

Using fundus perimetry with an SLO, Remky et al.(1996a) detected smaller scotoma depths (0.1–5 dB)than those found in the above-mentioned studies. Thiswas probably due to differences in experimental condi-tions (e.g. test strategy, stimulus colour).

Smaller (12%-) targets used in this study were usuallycharacterized not only by more pronounced angiosco-toma depths but also by a larger overall spread thanthat of the larger (32%-) test points. Therefore bothquantities—i.e. scotoma depth and spread parame-ter—were incorporated into the results via signal-to-noise ratio. The present study showed no relevantinfluence of target size on this ratio (see Fig. 8).

To the authors’ knowledge, this study was the first touse bright and dark stimuli for the detection of an-gioscotoma with a single campimetric device. The mon-itors used in this study were far superior to

Page 11: Angioscotoma detection with fundus-oriented perimetry A ... · horizontal black and gray lines show the methodologically achievable minimum of differential light sensitivity for black

U. Schiefer et al. / Vision Research 39 (1999) 1897–1909 1907

Fig. 7. General uncertainty in estimation (noise) s as a function of scotoma depth (h). Examination of 13 normal subjects with bright and darkstimuli of various diameters using the ‘one-step estimation’ procedure (see also Fig. 4).

conventional illumination techniques in generating testpoints which were darker than the surrounding back-ground (Aulhorn, 1964). However, careful calibrationwas required in order to guarantee reproducible homo-geneous background and stimulus luminance conditions(Dietrich et al., 1996).

The Tubingen Computer Campimeter (TCC) usessymmetric dB scales for bright and dark stimuli (atleast for luminance differences [DL ] between 910 cdm−2). For purposes of comparison, the luminance val-ues both bright and dark test points in this study weretherefore related to the background luminance—i.e. 10cd m−2. The reduced dynamic range of the dark stimulimay have impaired the detection and mathematicalcharacterization of angioscotomas: adding 10 cd m−2

to the background luminance level led to a compara-tively small increment in the luminance scale for brightstimuli, whereas subtracting the same value producedthe ‘darkest possible’ (absolutely black) value on thedark stimulus scale. This was especially true for thesmall, dark test points, which were presented withhigher contrasts (i.e. ‘darker’) than their counterpartswith greater diameters. This phenomenon can beviewed as a disturbing factor in both the single-threshold as well as the one-step angioscotoma estima-tion procedures.

Identical contrast thresholds for incremental anddecremental stimuli are to be expected in the absence ofstray light, whereas contrast thresholds for dark stimuliin the presence of stray light should be lower than thosefor bright ones. This study showed no relevant differ-ences between black and white stimuli on rates of

angioscotoma detection or signal-to-noise ratios. Thiswas to be expected, since scatter is negligible in the eyesof healthy young individuals and the intensity of straylight decreases with the distance from the visual axis(Miller & Benedek, 1973).

Cohn (1974) proposed that increment thresholdsshould exceed decrement thresholds due to assym-metries of the increment and decrement Poisson-distri-butions. However, these differences decline asbackground luminance is increased; this is especiallytrue for the photopic conditions of the experimentsdescribed here.

Mutlukan (1993, 1994a,b, 1995) was more successfulin detecting glaucomatous visual field defects with darktest points than with bright targets of similar size.However, in contrast to the present study, he useddifferent devices to present both dark and brightstimuli.

The questions addressed in Section 1 can therefore beanswered as follows:

(1) Smaller targets usually lead to more pronouncedangioscotoma depths but also to greater overall spreadthan that of larger test points. Angioscotoma widthswere not significantly influenced by target size.

(2) Signal-to-noise ratios showed no consistent de-pendence on target size and target luminance (i.e. incre-mental vs. decremental stimuli).

(3) Angioscotoma locations correspond well withthose of the causative morphological structure (i.e.retinal vessel). Angioscotoma width was usually some-what greater than the estimated diameter of thecausative retinal vessel.

Page 12: Angioscotoma detection with fundus-oriented perimetry A ... · horizontal black and gray lines show the methodologically achievable minimum of differential light sensitivity for black

U. Schiefer et al. / Vision Research 39 (1999) 1897–19091908

Fig. 8. Signal-to-noise ratio h/s in 13 normal subjects with bright and dark stimuli of various diameters using the ‘one-step estimation procedure’(see also Fig. 4).

5. Conclusions

Fundus-oriented perimetry proved to be an effectiveprocedure for detecting even minute visual field defectssuch as angioscotomas.

Neither variation of stimulus size nor luminancecharacteristics (increment vs. decrement) showed clearadvantages in respect to detection rates or signal-to-noise ratios for this type of scotoma.

Acknowledgements

The authors are indebted to Professor Dr. E. Plies,Institute of Applied Physics, Dr. A. Kurtenbach, De-partment of Pathophysiology of Vision and Neuro-Ophthalmology, Tubingen, and A. Remky, Departmentof Ophthalmology, RWTH Aachen, for their construc-tive criticism of the manuscript and T. Rice, M.A.,Tubingen, for his ‘linguistic help’. They would also liketo thank Miss K. Koller and Miss R. Hofer for theirsupport in preparing the graphics. The authors alsowish to thank the two anonymous referees for theirconstructive support.

References

Abe, T. (1968). Study on angioscotomas in glaucomatous eyes. ActaSocietatis Ophthalmologica Japan, 70, 59–66.

Aulhorn, E. (1964). U8 ber die Beziehung zwischen Lichtsinn undSehscharfe. Graefes Archi6e for Clinical and Experimental Oph-thalmology, 167, 4–74.

Baumgartner, G., Bornschein, H., Hanitzsch, R., Jung, R., Kornhu-ber, H.H., Rentschler, I., Schober, H., & Thoden, U. (1978).Sehen. Sinnesphysiologie III. Munchen, Wien, Baltimore: Urban& Schwarzenberg.

Bek, T., & Lund-Andersen, H. (1989). The influence of stimulus sizeon perimetric detection of small scotomata. Graefes Archi6e forClinical and Experimental Ophthalmology, 227, 531–534.

Bek, T., & Lund-Andersen, H. (1990). Accurate superimposition ofperimetry data onto fundus photographs. Acta OphthalmologicaCopenhagen, 68, 11–18.

Benda, N., Schiefer, U., & Dietrich, T. J. (1996). How to describeangioscotomata. German Journal Ophthalmology, 5 (Suppl. 1),127.

Bechetoille, A. (1996). Vascular risk factors in glaucoma. CurrentOpinion in Ophthalmology, 7, 39–43.

Chauhan, B. C., Henson, D. B., & Hobley, A. J. (1988). Clusteranalysis in visual field quantification. Documenta Ophthalmolog-ica, 69, 25–39.

Cibis, P. A. (1965). Lokaladaptation. Bericht uber die Zusammenkunftder Deutschen Ophthalmologischen Gesellschaft, 66, 193–202.

Cohn, T. E. (1974). A new hypothesis to explain why the incrementthreshold exceeds the decrement threshold. Vision Research, 14,1277–1279.

Dashevsky, A. I. (1938). Clinical angioscotometry—a new methodwith the use of different contrast test objects. Archi6es of Ophthal-mology, 19, 334–353.

Dietrich, T. J., Selig, B., Friedrich, M., Benda, N., & Schiefer, U.(1996). Calibration routines for video display units for perimetricexaminations. German Journal Ophthalmology, 5 (Suppl. 1), 125.

Eizenman, M., Trope, G. E., Fortinsky, M., & Murphy, P. H. (1992).Stability of fixation in healthy subjects during automated perime-try. Canadian Journal of Ophthalmology, 27, 336–340.

Fitzke, F. W., & McNaught, A. I. (1994). The diagnosis of visual fieldprogression in glaucoma. Current Science, 5, 110–115.

Goldmann, H. (1947). Beitrag zur Angioskotometrie. Ophthalmolog-ica, 114, 147–158.

Hudson, H. L., Frambach, D. A., & Lopez, P. F. (1995). Relation ofthe functional and structural fundus changes after submacularsurgery for neovascular age-related macular degeneration. BritishJournal of Ophthalmology, 79, 417–423.

Page 13: Angioscotoma detection with fundus-oriented perimetry A ... · horizontal black and gray lines show the methodologically achievable minimum of differential light sensitivity for black

U. Schiefer et al. / Vision Research 39 (1999) 1897–1909 1909

Haberlin, H., Funkhouser, A. T., & Fankhauser, F. (1983). Angiosco-toma: Preliminary results using the new spatially adaptive programSAPRO. In E. L. Greve & A. Heijl, Fifth International Visual FieldSymposium. The Hague: Dr W. Junk Publishers, 337–343.

Langerhorst, C. T., Carenini, L. L., Bakker, D., & De Bie-Raakman,M. A. C. (1997). Measurements for description of very earlyglaucomatous field defects. In M. Wall & A. Heijl, Perimetry update1996/1997. Amsterdam: Kugler, 67–73.

Littmann, H. (1988). Zur Bestimmung der wahren Grosse einesObjektes auf dem Hintergrund eines lebenden Auges. KlinischeMonatsblatter fur Augenheilkunde, 192, 66–67.

Lutz, S., Dietrich, T. J., Benda, N., Selig, B., Schiefer, U., & Daum,I. (1996). Computer perimetry using different techniques of inquir-ing answers. German Journal Ophthalmology, 5 (Suppl. 1), 125.

Menage, M. (1997). Medical therapy in chronic open angle glaucoma.CME Journal of Ophthalmolgy, 1, 12–13.

Miglior, S., Rossetti, L., Brigatti, L., Bujtar, E., & Orzalesi, N. (1994).Reproducibility of retinal nerve fiber layer evaluation by dynamicscanning laser ophthalmoscopy. American Journal of Ophthalmol-ogy, 118, 16–23.

Miller, D., & Benedek, G. (1973). Intraocular light scattering.Springfield, IL, USA: Charles C Thomas, 1–121.

Mutlukan, E. (1993). Glaucomatous optic neuropathy causes sensitiv-ity loss to light offsets in the visual field. NeuroReport, 4, 1159–1162.

Mutlukan, E. (1994). A comparison of automated static dark stimuliwith the Humphrey STATPAC program in glaucomatous visualfield loss. British Journal of Ophthalmology, 78, 175–184.

Mutlukan, E. (1994). Detection sensitivity to light offsets is abnormalin glaucomatous visual field. Documenta Ophthalmologica, 87,245–264.

Mutlukan, E. (1995). Glaucomatous visual field loss detected bythreshold light offset stimuli. In R. P. Mills & M. Wall, PerimetryUpdate 1994/1995. Amsterdam: Kugler, 147.

Nishida, Y., & Kani, K. (1997). A new fundus perimeter by which thetarget can automatically pursue eye movement. In M. Wall & A.Heijl, Perimetry Update 1996/1997. Amsterdam: Kugler Publica-tions, 75–79.

Orzalesi, N., Miglior, S., Lonati, C., & Rossetti, L. (1998). Mi-croperimetry of localized retional nerve fiber layer defects. VisionResearch, 38, 763–771.

Remky, A., Beausencourt, E., & Elsner, A. E. (1996a). Angioscotome-try with the scanning laser ophthalmoscope—Comparison of theeffect of different wavelengths. In6estigati6e Ophthalmoly & VisualScience, 37, 2350–2355.

Remky, A., Elsner, A. E., Morandi, A., & Beausencourt, E. (1996b).Nasal-temporal asymmetry in the central field assessed by blue-on-yellow perimetry with a SLO. XII IPS Meeting, Wurzburg, Program& Abstracts, 56 (abstract).

Rohrschneider, K., Becker, M., Kruse, F. E., Fendrich, T., & Volcker,H. E. (1995a). Stability of fixation: results of fundus-controlledexamination using the scanning laser ophthalmoscope. GermanJournal Ophthalmology, 4, 197–202.

Rohrschneider, K., Fendrich, T., Becker, M., & Volcker, H. E. (1995b).Zum Einsatz einer kinetischen funduskontrollierten Perimetrie mitdem Scanning Laser Ophthalmoskop. Ophthalmologica, 92 (Suppl.1), 52.

Rohrschneider, K., Gluck, R., Burk, R. O. W., Kruse, F. E., & Volcker,H. E. (1997). Macular sparing in patients with hemianopsia -re-evaluated using static and kinetic fundus perimetry. In M. Wall& A. Heijl, Perimetry update 1996/1997. Amsterdam: Kugler,377–385.

Safran, A. B., Halfon, A., Safran, E., & Mermoud, C. (1995).Angioscotomata and morphological features of related vessels inautomated perimetry. British Journal of Ophthalmology, 79, 118–124.

Schiefer, U., Stercken-Sorrenti, G., Dietrich, T. J., Friedrich, M., &Benda, N. (1996a). Fundus-orientierte Perimetrie-Evaluation einesneuen Gesichtsfeld-Untersuchungsverfahrens bezuglich der Detek-tion von Angioskotomen. Klinische Monatsblatter fur Augen-heilkunde, 209, 62–71.

Schiefer, U., Stercken-Sorrenti, G., Dietrich, T. J., Selig, B., Benda, N.,& Friedrich, M. (1996b). Improvement of diagnostic efficiency byfundus - oriented perimetry. German Journal Ophthalmology, 5(Suppl. 1), 126.

Schiefer, U., & Witte, A. (1996). Patent: Perimetrisches Unter-suchungsverfahren–Fundusgestutzte Perimetrie II. DeutschesPatentamt Munchen, Az196 21 9612, 1–15.

Schneider, U., Kuck, H., Inhoffen, W., & Kreissig, I. (1993). Fun-duskontrollierte Mikroperimetrie mit dem Scanning-Laser-Oph-thalmoskop bei Makulaerkrankungen. Klinische Monatsblatter furAugenheilkunde, 203, 212–218.

Sjaarda, R. N., Frank, D. A., Glaser, B. M., Thompson, J. T., &Murphy, R. P. (1993). Resolution of an absolute scotoma andimprovement of relative scotomata after successful macular holesurgery. American Journal of Ophthalmology, 116, 129–139.

Spillmann, L., & Kurtenbach, A. (1992). Dynamic noise backgroundsfacilitate target fading. Vision Research, 32, 1941–1946.

Straub, W., Kroll, P., & Kuchle, H. J. (1995). Augenarztliche Unter-suchungsmethoden. Stuttgart: Enke.

Sturmer, J. (1993). Fundusperimetrie. In B. Gloor, Perimetrie mitbesonderer Berucksichtigung der Automatischen Perimetrie. Stutt-gart: Enke, 149–156.

Sunness, J. S., Schuchard, R. A., Shen, N., Rubin, G. S., Dagnelie, G.,& Haselwood, D. M. (1995). Landmark-driven fundus perimetryusing the scanning laser ophthalmoscope. In6estigati6e Ophthalmoly& Visual Science, 36, 1863–1874.

Timberlake, G. T., Mainster, M. A., Webb, R. H., Hughes, G. W., &Trempe, C. L. (1982). Retinal localization of scotomata by scanninglaser ophthalmoscopy. In6estigati6e Ophthalmology & Visual Sci-ence, 22, 91–97.

Troxler, D. (1804). U8 ber das Verschwinden gegebener Gegenstandeinnerhalb unseres Gesichtskreises. In K. Himly & J. A. Schmidt,Ophthalmologische Bibliothek. Jena: Springer, 1–53.

Tuulonen, A., Lehtola, J., & Airaksinen, P. J. (1993). Nerve fiber layerdefects with normal visual fields. Do normal optic disc and normalvisual field indicate absence of glaucomatous abnormality? Oph-thalmology, 100, 587–597.

Weekers, R., & Humblet, M. (1945). L’Angioscotome physiologique.Ophthalmologica, 110, 43–59.

Welt, M. (1945). Etude sur les rapports entre les dimensions de la tacheaveugle de Mariotte et des angioscotomes, et la tension arterielleretinienne. Ophthalmologica, 109, 137–158.

Zalta, A. H., & Burchfield, J. C. (1990). Detecting early glaucomatousfield defects with the size I stimulus and Statpac. British Journal ofOphthalmology, 74, 289–293.

Zulauf, M. (1988). Beitrag zur Angioskotometrie: Stimulusgrosse undProgrammwahl. Klinische Monatsblatter fur Augenheilkunde, 192,613–618.

Zulauf, M. (1990). Pindolol und Timolol: Short term influence onangioscotomas and static brightness contrast sensitivity. Ophthal-mologica, 201, 37–44.

Zulauf, M. (1990). Quantification of angioscotomas. Ophthalmologica,200, 203–209.

.