automated static perimeter/adaptometer light emitting diodes · in this paper we describe a static...

British Journal ofOphthalmology, 1983, 67, 431-442

An automated static perimeter/adaptometer usinglight emitting diodesW, ERNST, D. J. FAULKNER, C. R. HOGG, D. J. POWELL, G. B. ARDEN, AND VAEGAN

From the Department of Visual Science, Institute of Ophthalmology, London, and Electrodiagnostic Clinic,Moorfields Eye Hospital, London

SUMMARY An automated static perimeter/adaptometer is described which measures thresholdswith lights of 2 wavelengths. The instrument uses light-emitting diodes to produce the stimuli and iscontrolled by a small computer, making it very suitable for clinical testing of large numbers ofpatients. The use of 2 LEDs with different peak emission wavelengths (530 and 660 nm) permits anassessment of the relative state of rod and cone mechanisms in a particular region of the retinaeither during dark adaptation or when the eye is fully dark adapted.

A knowledge of the state of rod function relative tocone function in a particular part of the visual field isvaluable in the diagnosis and management of retinaldisorders and essential in any research on the natureof a particular disorder. Static perimetry on the darkadapted eye can provide much of the informationrequired, and a number of instruments, notably theTubinger perimeter,' have been designed to meet thisneed. However, static perimetry is not routinely usedas a clinical technique on large numbers of patientsbecause of the practical problems it presents:measurements take a long time to execute; the pro-cedures are involved and require the full attention ofa trained operator; most static perimeters areelaborate and expensive; analysis of results is complexand may need extra time and effort after the actualmeasurement. In this paper we describe a staticperimeter/adaptometer which we have built fromlow-cost components; its automation proved straight-forward because light-emitting diodes (LEDs) wereused as sources for the stimuli. Computer controleases the burden on the operator and permitsstandardisation of procedure, especially in the pacingof a series of measurements and in the criteria usedfor establishing patient thresholds. Further, resultscan be analysed immediately after measurementshave been completed. These improvements haveenabled us to undertake a large-scale study of staticfields in over 200 patients suffering from retinitispigmentosa which will be described in detail in sub-sequent papers.Correspondence to Dr W. Ernst, Department of Visual Science,Institute of Ophthalmology, Judd Street, London WCIH 9QS.

Static perimeters and adaptometers often use awhite light stimulus, but in the examination of an areaof the visual field where thresholds are elevated it maybe important to establish whether it is primarily rodfunction which is affected or whether there is a loss inboth rod and cone systems. Measurements with whitelight cannot provide this information. To identifywhich mechanism determines the results, the bestapproach is to vary the spectral composition of thethreshold stimulus. Spectral sensitivity plots can bederived from such measurements2 and compared withthe scotopic function if the rod pignent is presumedresponsible for light absorption, or the photopic oneif the cone pigments are at work, or a composite curveif rod and cone sensitivities are nearly equal. How-ever, this approach requires the collection of a verylarge amount of data from each patient. A morepractical method is to vary the stimulus between only2 colours: Massof and Finkelstein34 have shown thatsatisfactory tagging of threshold data can be achievedwhen blue-green and deep red stimuli are used. Inthis paper we describe how we have adapted the2-colour idea for our perimeter/adaptometer anddemonstrate how our analysis of the results candistinguish between a selective disturbance of rodvision and a nonselective disturbance involving bothrods and cones.

Materials and methods

PERIMETER CONSTRUCTIONA modified Lister perimeter5 forms the basis of theinstrument we have built. The novel feature is that

431

on February 25, 2020 by guest. P

rotected by copyright.http://bjo.bm

j.com/

Br J O

phthalmol: first published as 10.1136/bjo.67.7.431 on 1 July 1983. D

ownloaded from

http://bjo.bmj.com/

W. Ernst, D. J. Faulkner, C. R. Hogg, D. J. Powell, G. B. Arden, and Vaegan

the targets of this perimeter have been replaced withlight emitting diodes (LEDs) used in such a way as togive a working range of 5 log units of light intensitychange.

Fig. 1 is a diagram showing the main features of theperimeter. A single red LED has been placed at thecentre of the arm to act as a fixation point for thedetection of eccentric stimuli by the observer. Thetest stimuli are mounted on the target carriage. Therecording equipment normally found at the back hasbeen replaced by two reduction gear systems drivenby stepping motors. The first drives the targetcarriage along the arm to a given eccentricity. Thesecond rotates the arm so that the target is in theappropriate meridian. The observer's head is placedon a chin rest which can be adjusted so that the pupilof the eye being tested is at the centre of the arcdescribed by the arm and the LEDs point at thepupil.The test stimuli are formed from 2 adjacent LEDs,

namely, Stanley ESBG5501 (green) and ESBR5501(red) fitted into a small metal can. Each LED sub-tends an angle of 0.90 at the observer's eye. An Ilford604 filter covers the greenLED and 2 Ilford 608 filtersthe red one. Fig. 2 shows the spectral output of theLEDs unfiltered (dotted lines) and filtered (solidlines). It can be seen that the dominant wavelength ofthe green light is about 530 nm and that of the redlight about 660 nm.

LUMINANCE CONTROLThe continuous current required to operate an LEDat maximum luminance is about 20 to 60 mA at a

:Motors

/ -Fixation LED

/,'I'f

/

/

Fig 1 Modification ofthe Lister perimeter. The observer'sright eye is shown being tested and the target carriage is at an

eccentricity of260 on theO (horizontal, temporal) meridian.The target can contains a red and a green LED (not shown).

100 r

c0

50

(ap

A'L

500 550 600 650 700Wavelength

Fig 2 Relative light outputfrom Stanley ESBG5501 andESBR5501 LEDs. Dotted lines: LEDs unfiltered. Solidlines: ESBG outputpassed through Ilford 604filter, ESBRoutput passed through 2 Ilford 608filters.

forward voltage of 1-6 to 2 V. The light intensity is anonlinear function of current,6 approximating to apower law with an exponent of 1-3 to 1-5. While thisnonlinearity may be unimportant in some applica-tions, and can be compensated for over an intensityrange of perhaps 1 to 2 log units, it is not reallypractical to control intensity over the wide rangerequired for a perimeter simply by using variable-amplitude current drive. A better technique is tooperate the LED in a pulsed mode and vary the pulserepetition rate. Since the pulse response of an LED isfast, with typical rise and fall times of 100 ns, pulsewidths of less than 1 ,s and repetition frequencies ofover 1 MHz are possible. At repetition rates abovethe critical flicker fusion frequency the eye does notperceive individual pulses but integrates them to givea time average. With a fixed pulse width the lightintensity is a linear function of pulse repetition rate.We have therefore used repetition rate for coarseintensity control and current drive for fine steps.

Fig. 3 shows a block diagram of the system forcontrolling the output of the LEDs. Each LED isdriven by its own current source (A) which can beprogrammed to produce 2 to 20 mA in preset steps of0 1 log units. This provides fine intensity control over0 9 log units. For most of the time transistor (Q) is on,shunting current away from the LED and turning itoff. During the test presentation 1 ,us-wide pulsesfrom the monostable (B) turn off transistor (Q) andturn on the LED. The pulse repetition rate is derivedfrom clock 1 (C) which runs at 350 kHz. A decadedivider chain (D) produces pulses at rates from 350kHz to 35 Hz and pulses at one of these frequenciespass through the data selector (E) to trigger themonostable (B). This provides a further 4 log units of

432

/



j.com/

Br J O


ownloaded from

http://bjo.bmj.com/

An i uladaptometer using light emitting diodes

CLOCK 2Hz BCD COUNTER DISPLAY REFERENCEG H i VOLTAGES

Fig 3 Block diagram ofcomponents used to drive LEDs. For explanation see text.

intensity control in 1 log unit steps, giving a totalrange of 4-9 log units.

Dividing the intensity control into coarse and finesections has the useful side effect that only 2 binary-coded decimal (BCD) digits are required to programintensity directly in log units. The perimeter can beplaced under manual or computer control by dataselector (F). Manual control is achieved by usingclock 2 (G) to increment or decrement a BCDup/down counter (H) with a range of0 to 49., A directreadout of the intensity setting in log units is providedby a 2 digit display (J). The timing of the test presen-tation is set at 200ms on, 800ms offby clock 3 (K).

Since each LED in the array is driven by its owncurrent source, a red or green stimulus can be selec-ted simply by removing the drive pulses from theunwanted section. This gating action is provided bythe tri-state buffers (L) which also interface theCMOS control circuitry to the current sources.The maximum luminance of the filtered green

LED, corresponding to a reading of4 9 on the display(J), is 0-9 log cd m-2 and that of the filtered red one is2-2 log cd m-2. This results in normal observers givingmean log threshold readings of 1-2 and 1-4 on display(J) for the green and red stimuli, respectively, whentested parafoveally. Thus the stimuli can always bereduced below the threshold luminance of even themost sensitive observer tested in his/her most sensitiveregion. Further, the differences between the normalreadings and the maximum available reading (4 9) are

sufficiently large for sensible quantification ofthreshold elevations in patients with retinal disorders.

AUTOMATIONThe perimeter has been linked to a minicomputer(Plessey Micro 1) housed outside the darkened areain which the perimeter is placed. Before measure-ments are made the computer calibrates the rotaryand eccentric movements of the stimulus by findingthe number of pulses required by the stepping motorsto move the perimeter arm and target carriagebetween the permitted extremes. The operator entersbasic patient details, including an identifying numberused to name a file that will contain the results of thistest. Clinical notes may be entered and stored withthe test results. The operator then calls for an'instruction' file which contains all the informationneeded for the computer to test the patient in adefined manner-that is, the meridians to be investi-gated in appropriate order and the eccentricities andcolours for which thresholds are to be measured.When the patient is ready to begin the test, he/she

presses a start button. On this signal the computersets the intensity of the test stimulus to the lowestvalue possible and then increases it at a constant rate.As soon as the stimulus is visible, the patient presses aresponse key. The intensity value is stored, and theintensity is then decreased by an amount which variesrandomly between 07 and 1-3 log units. This pro-cedure is repeated until the standard error of the last 4

automated static Derimeter,Iw.1- -r 433



j.com/

Br J O


ownloaded from

http://bjo.bmj.com/


10

-c-3

.!3 2

4h

51Fig. 4 Dark adaptation curvesobtainedfrom a normal observer (Iin Table 2) detecting the green andred stimuli at an eccentricity of250on the nasal, horizontal meridian ofhis right eye (1800). Open circles:red (660 nm) stimulus; closedcircles: green (530 nm) stimulus.Ordinate: log relative thresholdenergy. All values are referred to theenergy ofthe green stimulus whendisplay (J)-see Fig. 3-shows azero reading. The green stimulusthen has a luminance of -4 log cdm-2. Forfull explanation see text.

O0L0

1_

0 00

0 0, c

ti t2

10 20 30 40Time(nin)

responses is less than or equal to 041 log unit. Such a

procedure is adopted to achieve a consistent set ofdata from which to estimate log threshold. As a

precaution previous thresholds and individual res-

ponse values are displayed graphically on a visualdisplay unit for the operator, who if necessary can

halt measurements to reinstruct the patient, slowdown the rate of intensity change, or alter thestatistical criterion used by the computer to define thepatient's threshold. Once a series of readings are

complete, the computer indicates this to the patientby a tone, and if the patient is ready to continue to thenext set of measurements he presses the start buttonagain. Measurements are organised so that thestimulus moves to successively greater eccentricitiesalong a meridian. When a meridian is complete, theoperator has the choice of continuing to anothermeridian, repeating the same meridian, or quittingthe test. A patient is usually tested at between 30 and40 locations in the visual field for the 2 stimuli (seeFigs. 9 and 10). The procedure lasts about 25minutes.The results of all tests are stored immediately on

the computer disc. A variety ofprograms are availablefor analysing the data, including averaging resultsfrom a number of observers so that groups can becompared. The most commonly used method ofanalysis is to obtain plots of the patient's thresholdelevation compared with normal values as a functionof eccentricity along individual meridians.

PROCEDURE WITH PATIENTS

Before the test the patients are shown the apparatusand given the opportunity to practise using the startand response keys. They are told that after pressingthe start key they must look continuously at the steadyred light and press the response key as soon as theybecome aware of a pulsing light, however faint, some-where 'in the corner of the eye'. The patients are

always informed of the particular meridian selectedfor testing. In the measurement of central thresholds,when the pulsing target obscures the fixation light,the patients are instructed to look directly for thepulsing light on its first appearance and to keep theirgaze fixed once the target has been located. Theoperator usually remains with a patient until he issatisfied that the patient understands the nature of thetest. If the patient seems slow to respond to theappearance of the test light, the operator slows downthe rate at which the test light intensity is ramped up.Patients are given frequent reminders of theimportance of keeping their gaze steady on thefixation light. Rest periods of several minutes are

arranged for patients during the change-over fromthe investigation of one meridian to another.The patients' pupils are always dilated for the test,

usually with a mixture of phenylephrine and cyclo-pentolate. A minimum period of 20 minutes of darkadaptation is allowed before the first measurement.No correction is attempted for any defects in visualacuity.

00

co Ow 9 00000

t35 650 60 70

434

6r



j.com/

Br J O


ownloaded from

http://bjo.bmj.com/

An automated static perimeterladaptometer using light emitting diodes

10- t1 t2

8H

56

c

.-j

2

*I IW500 700 500 700

Wavelength (nm)

"sIsrF

500 700

Fig. 5 Photopic (solid line) and scotopic (dashed line)spectral threshold functions describing the data obtained at 3instants (t,, t2, and t3) ofthe dark adaptation experiment ofFig. 4. Ordinate asfor Fig. 4. LettersA, C, and Eindicate theposition ofthe 660nm dark adaptation curve at the3 instants;letters B, D, and Findicate theposition ofthe530nm curve; yindicates the log threshold minimum ofthephotopicfunction(at about 550 nm) which does not change between theinstants; Y,, Y2, and Y3 represent the successive minima ofthe scotopic function (at about 500 nm) as rod darkadaptation proceeds.

THEORY

Fig. 4 shows 2 dark adaptation curves obtained from anormal subject after a flash bleach covering the 0 9°test area which was centred at 250 on the horizontalnasal meridian of the right eye. Measurements werebegun after an initial cone dark adaptation phase.Open circles are measurements made with a red testlight (660 mm), while closed ones are those with agreen test light (530 nm). At time t, both test lightswere detected by the cone mechanism to give logthreshold values of A and B. By time t2 the rodmechanism had recovered sufficiently to be involvedin the detection of the green light and give a thresholdvalue of D. The detection of the red light was stillmediated by the cones, and the log threshold, C, wasequal to A. By t, the rod mechanism had reached itsmaximum sensitivity. There is a slight drop inthreshold for the red test light, and it is possible thatthe measurement, E, was mediated by the rodmechanism, which clearly was also responsible for thevalue, F, obtained with the green light.The relationship of these threshold changes to the

photopic and scotopic spectral sensitivity functions isshown in Fig. 5, where the reciprocals of CIEphotopic and scotopic spectral sensitivity coefficients7have been transformed into logarithmic ratios,allowing a direct comparison with Fig. 4; the curvesare thus inverted log sensitivity functions. The figureshows the functions for the 3 instants of darkadaptation, t1, t2, and t3, labelled in Fig. 4. From Fig.5 it can be seen that the difference between logthreshold values A and B is characteristic of thephotopic spectral threshold function, while thedifference between E and F is characteristic of thescotopic one. A difference such as C-D (at t2) whichis greater than A-B but smaller than E-F indicatesthat the green light is being detected by the rods andthe red one by the cones.

Fig. 5 shows that during dark adaptation thespectral threshold functions of rod and conemechanisms are moving relative to each other. Aftercone adaptation the minimum log threshold of thecone function (at about 550 nm) equals y. At t1 therod function lies somewhere above the cone. Its exactposition is unknown, but the curve has been placed togive a 530 nm reading, which corresponds to theextrapolated value of the rod dark adaptation curveat t1 (Fig. 4). At t2 the minimum log threshold of therod function (at about 500 nm) equals Y2; at t3, whenthe eye is dark adapted, it equals Y3. The differencebetween the log threshold minima for the twofunctions, y-Y3, is dependent on the characteristicsof the test target, also on the relative numbers of rodsand cones in the region being tested, on their relativeefficiency in absorbing the incident light, and on thetemporal and spatial characteristics of the neuralnetworks controlled by each type of receptor. Thedifference will determine whether the value E-F isequal to the value expected from a scotopic thresholdfunction as shown in Fig. 5 or whether it is smaller.The latter condition results either from an overlapbetween the scotopic and photopic functions or, inthe extreme case, the failure of the scotopic functionto move below the photopic one.Consider nowwhat may happen in the dark adapted

eye if there is a disturbance of rod and/or cone visionin the region due to disease. If both receptormechanisms are affected and the cone spectralthreshold function moves up to the same or to agreater extent than the rod one, then the latter willcontinue to describe measurements at 530 and 660nm. Hence the log threshold elevations relative tonormal, (E'-E) and (F'-F), will be equal asillustrated in Fig. 6. However, if the rod functionmoves up while the cone function remains un-disturbed, or if both functions move up but the rodone to a greater extent, then the resulting situationbecomes analogous to that observed in a normal

435

OL_



j.com/

Br J O


ownloaded from

http://bjo.bmj.com/

W. Emst, D. J. Faulkner, C. R. Hogg, D. J. Powell, G. B. Arden, and Vaegan

II II I

500 700I I I I I

500 700

seen in the rod spectral log threshold function for the2 wavelengths less the difference seen in the conefunction. It is a measure of the Purkinje shift inmoving from one spectral threshold function to theother and is dependent only on the characteristics ofthe 2 functions. The last statement, however, doesrequire qualification in one respect. In some regionsof the normal retina the state of affairs shown for t2 inFig. 4 may represent the final state of dark adaptation,that is, the 2 spectral functions do not separatesufficiently for the scotopic function to describe thefinal measurement with the red test light. In thesecircumstances the maximum value of the index will bereduced by an amount which can be calculated if theseparation of the functions is known.To summarise, a disturbance in rod function in a

particular region can be identified by a log thresholdelevation for the green test light. Figs. 4 and 5 showthat elevations up to about 2 log units are measures ofthe loss of rod function. The interpretation of greaterelevations requires knowledge of what is happeningto the cones. If there is little deviation from normal inthe red measurement, the cones in the region are

Wavelength (nm)Fig. 6 Normal scotopic and photopic spectral thresholdfunctions in the dark adapted eye (left panel) and bothfunctions elevated to the,same extent by disease (rightpanel).Ordinate as for Fig. 4. For explanation see text.

observer at various instants before dark adaptation iscomplete. From Fig. 4 it can be seen that (D-F) isgreater than (C-E) or that (B-F) is greater than(A-E): there is a greater log threshold elevation forthe green light than for the red one.

In our presentation of perimetry results frompatients tested at a series of locations we show in 2separate panels their log threshold elevations for the 2stimuli relative to normal values. The normal valuesare means obtained from 10 to 30 individuals. We alsodisplay in a third panel the difference between the logthreshold elevations (green minus red) which we havecalled the green-red index. When the index is zero, itcan be seen from Fig. 6 that the disorder is affectingthe cones to the same or greater extent than the rods.Consider now the other extreme possibility, namely,rod function is so much more affected that thepatient's thresholds for both green and red light are

mediated by cone vision. Under these conditions theindex assumes a maximum positive value. Here it maybe helpful to compare the patient with the normalobserver at t, during dark adaptation and to refer toFig. 4. The log threshold elevation for green is (B-F)and for red is (A-E). The index is therefore(B-F)-(A-E) or (E-F)-(A-B). Thus themaximum value of the index represents the difference

0~~ ~ ~

2

0-

L I

0 20 40 60 80Eccentricity(degrees)

Fig. 7 Data from an RPpatient examined on the LEDperimeter (filled circles, solid line) compared with datafromthesame patient tested on a Tubingerperimeter (open circles,dashed line). Dotted line represents the averageperformanceofa group ofnormal observers examined on the LEDperimeter under the same conditions as the patient.Dominant wavelength for red stimulus: 626 nm for the LEDperimeter; 600 nm for the Tubinger. Forfull description see

text. A zero value on the log threshold axis corresponds to a

luminance of -4 9 log cd m-2 for the red LED.

8

6 _-o0

Cf)a)

0).0-J

2_

OL

436

4_



j.com/

Br J O


ownloaded from

http://bjo.bmj.com/

An automated static perimeterladaptometer using light emitting diodes

being spared by the disease. If, however, there is anelevation of the log threshold reading for the redlight, the index needs to be examined to determinewhether, on the one hand, the disturbance is affectingcone vision to at least the same or to a greater extentthan rod vision or, on the other, it is rod vision whichis being disturbed to a greater extent. In the formercase the index will be close to zero; in the latter it willhave a significant positive value. Provided the index isless than its maximum possible value for the givenlocation, the elevation in the green measures rod loss.However, when the index maximum is reached, thedetection of both red and green stimuli is mediated bythe cones, and only a lower limit can be set on theextent of rod loss.

Results

RELIABILITY OF THE LEDPERIMETER/ADAPTOMETERThe Tubinger perimeter has been extensively used tomap complex rod abnormalities in retinitis pig-mentosa (RP).34 Fig. 7 shows thresholds for the rightdark adapted eye of a sufferer from the autosomaldominant form of RP. The open symbols joined bythe dotted line represent measurements made with aTubinger perimeter. The 150 meridian of the right eyewas chosen; the target subtended 1.10 at the patient'seye; the colour filter resulted in the stimulus lighthaving a dominant wavelength of 600 nm; theduration of the stimulus was 200 ms; thresholds weredetermined by the operator using an ascendingmethod of limits with single flash presentations.Seven weeks later the patient was tested with ourautomated LED perimeter. The same meridian wasexamined; the target was smaller, namely, 0.90; thedominant wavelength of the light output from theLED was 626 nm (in this experiment the prototypeperimeter was used-see 'Discussion'-and the LEDwas a Hewlett Packard 5082-4658 and not the StanleyESBR5501); the stimulus lasted 200 ms and wasrepeated every second; measurements were collectedautomatically, as described. The results are shown asthe filled symbols joined by the solid line. TheTubinger measurements have been scaled to equalthe LED ones at a 100 eccentricity. It can be seen thatboth types ofmeasurement gave rise to similar shapedprofiles of visual function along this meridian.

Fig. 8 shows the degree of reproducibility that canbe expected in data obtained from an RP patienttested under the same conditions over 2 sessions. Theopen symbols represent data obtained from the firstsession, while the filled symbols refer to data obtained2 months later. Circles represent measurements madeat the beginning of the session; squares are readingsrecorded after 2 other meridians were investigated.

5

4

2--J

O0

0

00

0

° 0

o 0

U

0

0

0 10 20 30Eccentricity (degrees)

Fig. 8 Threshold measurements on an RPpatient collectedduring2 sessions 2 months apart and at different times duringeach session. Circles: measurements at the beginning ofeachsession; squares: measurements made at the end ofeachsession; open symbols: measurements made at thefirstsession; closed symbols: measurements made 2 months later.Green stimulus; horizontal nasal meridian ofthe right eye. Azero log threshold reading corresponds to a luminance of-4 log cd m-2.

To check the validity of using the green-red indexto label rod and cone function we carried out a seriesof dark adaptation experiments of the type illustratedin Fig. 4 on a number of normal observers for a varietyof locations in the visual field. The results are sum-marised in Table 1. Column 1 identifies the observers.Column 2 identifies the locations in the visual field of

Table 1 Separation ofphotopic and scotopic spectralcurves and green-red indices for various retinal locations

1 2 3 4 5

Observer Eccentricity, y- Y3 Expected index Observed indexdegrees maximum maximum

I 5N 2-0 2-0 1-61 25N 2-0 2-0 1-82 5N 1-9 1-9 1-62 25N 2-3 2-3 1-93 5N 2-2 2-2 1-83 25N 2-1 2-1 2-34 10T 25 2-3 2-04 20T 30 2-3 2-14 40T 2-8 2-3 2-44 80T 2-7 2-3 2-0

N = nasal. T = temporal.

0

40 50

437



j.com/

Br J O


ownloaded from

http://bjo.bmj.com/


Fig. 9 Analysis ofRP data. Logthreshold elevation relative tonormalfor green stimulus (leftpanel) andfor red stimulus(middle panel), and the green-redindex (right panel) plotted as afunction ofeccentricity on 4meridians of the right eye (00, 315°,2250, and 1600). Patient: AD RPtype 1.

green4 - 0 * 0.

2 _ o

0

4-315 t tt

00- 2 -

o 4- 225°0 0

225004-

4 16

2 4 ---

0 40 80

red,tt ttt

000 t- 'le - -e- -4

t t

0

1

0 40 80

indextffttt t

0 0I

* - -- -

ft t t

I . . L0 40 80

the right eye. All locations are on the horizontalmeridian and are defined in degrees of eccenticity; Nrefers to nasal, T to temporal. The separation incolumn 3 refers to the distance between the minima ofthe rod and cone spectral threshold functions, y-Y3,as illustrated in Fig. 5. The values are derived frommeasurements with the green stimulus and from pub-lished data on the 2 functions. Column 4 gives theexpected index maxima derived from column 3. At660 nm the threshold for the photopic function waselevated by 1-2 log units relative to its minimum; thecorresponding value for the scotopic function was 3 5log units.7 Therefore, in theory, for the red lightmeasurements to be described by the scotopicfunction the separation between the minima of the 2functions must equal or exceed 2-3 log units for thegiven region of retina being considered. If that is thecase, the maximum theoretical value of the index can

be calculated from the 2 functions. It is given by(E-F)-(A-B) (see Fig. 4). This turns out to be 2-3log units. If, however, the separation between the 2functions falls short of the 2-3 value by an amount x

log units, then x must be subtracted from 2-3 to givethe expected index maximum. An empirical indexmaximum can also be calculated from each observer'sdata for a given location by subtracting the greenfrom the red log threshold values during the coneplateau of dark adaptation to make difference (1),

doing a similar subtraction for the period at the end ofdark adaptation to make difference (2), and finallysubtracting difference (2) from difference (1). Suchdouble differences are shown in column 5. In general,the empirical values were lower than the onescalculated from the 2 functions. Detailed examinationof the data shows that, whereas the CIE basedscotopic function gives a very good description of themeasurements made on the dark adapted eye, thephotopic function consistently underestimates the logthreshold value for the red test light by about 02 logunit. In the analysis of data we have accordinglydecreased the expected index maximum by 0-2 logunit.

EXAMPLES OF THE CLINICAL APPLICATIONS OFTHE LED PERIMETER/ADAPTOMETERFig. 9 illustrates the way we usually analyse our

patient data. The example shown is for a 12-year-oldboy suffering from the autosomal dominant form ofRP. Table 2 outlines how the plots are constructedfrom readings for the 00 meridian. Each of the meannormal values is subtracted from the patient value togive a log threshold elevation or green-red index.Asterisks in the table and upward arrows in the figureindicate that the patient was unable to see thestimulus at its maximum luminance. The 'green' and'red' plots suggest that the patient had a ring scotoma

0 --

00 . - -a&0

'*-. v.*

*!0 ---.

0

0

438



j.com/

Br J O


ownloaded from

http://bjo.bmj.com/

An automated static perimeter/adaptometer using light emitting diodes 439

Table 2 RPpatient data for O meridian ofthe right eye compared with normal

Eccentricity Patient Nonmal Elevation Patient Normnal Elevation Patient Normalgreen green green red red red green-red green-red

0 3-5 2-1 1-4 1-7 1-4 0-3 1-8 0-75 3-5 1-5 2-0 1-9 1-5 0-4 1-6 0-010 3-3 1-4 1-9 2-1 1-7 0-4 1-2 -0-320 3-5 1-7 1-8 1-9 2-0 -0-1 1-6 -0-325 3-7 1-5 2-2 2-0 2-0 0-0 1-7 -0-530 3-5 1-5 2-0 2-2 1-8 0-4 1-3 -0-335 3-7 1-6 2-1 * 2-0 * * -0-440 * 1-6 * * 1-9 * * -0-345 * 1-8 * * 2-0 * * -0-250 * 1-8 * 2-1 2-1 0-0 * -0-355 * 2-0 * * 2-3 * * -0-360 * 2-1 * * 2-3 * * -0-265 3-8 2-2 1-6 * 2-4 * * -0-270 3-7 2-3 1-4 2-2 2-5 -0-3 1-5 -0-2

developing. Examination of the index shows that foralmost all readings it is close to its maximum value,indicating that, even when the patient had someresidual rod function, rod thresholds lie well abovecone ones. At least in some regions cone function wasnear normal, since the 'red' readings were close to thezero line.

Fig. 10 shows a similar set of plots from anotherpatient with autosomal dominant RP. Here it is clearthat the pattern of disease is very different. There isevidence of a ring scotoma but in regions where the

log threshold elevation could be measured the indexwas close to zero, indicating a generalised disturbanceof rod and cone function and not a selective rod loss.The comparison between Figs. 9 and 10 illustrates theimportance of 2-colour static perimetry in differentialdiagnosis.

Fig. 11 demonstrates the importance ofthe 2-colourtechnique in the correct interpretation of adapto-metric results.These data are taken from a study of patients with

liver disorders (Kemp, Ernst, Walt, Sherlock, Lyness,

0green4- 00 ttt4~~~~~2-

4 315 t

r 2

0-

oUO 4L 2250

-4

8 '1600J4r

2k

O--

0 40 80

redtt

*0

0

00 00

t

0

0 0 0 _ _

- - -

i I I A A a A* *

0 40 80

index

,1 ......................0

.-9 0 - --.

.....................

0j . .

- a 0. 0-o - -

.........

* 40

0 __0-A.----

Fig. 10 Analysis ofRPdata. Seelegend to Fig. 9. Patient: AD RPtype 2.

Eccentricity (degrees)



j.com/

Br J O


ownloaded from

http://bjo.bmj.com/


7r

6k

Fig. 11 Dark adaptation dataobtainedfrom a patient sufferingfrom a liver disorder. Details as inFig. 4. Open symbols: red stimulus;closed symbols: green stimulus.

V

I-en4

5

0).133

000

0co

0 00 00 0 0 000 000 0 0 0

0 00 0 0 0

0Sm 0

000

* 0 0 0 00

* 0

00

0

0

0

00 0

0 0

0

0 0

0

0

0

0

0

21_

1'-II I I

0 10 20 30 40Time(min)

and Bird, unpublished) and can be compared with thenormal data illustrated in Fig. 4, since they were

obtained under very similar conditions. If themeasurements had been conducted with just thegreen stimulus, the interpretation might easily havebeen made that part of the curve represented roddark adaptation, albeit abnormal. However, theadditional readings with the red light show that thewhole of the dark adaptation displayed is due to thecones and that it is very significantly delayed relativeto normal.

DiscussionThe usefulness of measuring spectral sensitivity incases of RP was first appreciated by Zeavin andWald.8 Massof and Finkelstein34 demonstrated that,where large numbers of individuals and limitedtesting time were involved, a practical solution was torestrict measurements to 2 stimuli of differentcolours, one blue-green (500 nm) and the other deepred (650 nm). In more extensive spectral measure-ments on normal observers and RP patients, Massofet al.2 confirmed the validity of this approach. Wehave therefore designed our automated staticperimeter to have 2 spectrally distinct stimuli. LikeMassof and Finkelstein34 we have found at least 2distinct sub-types of autosomal dominant RP. Inone-classified as type 1 by Massof and Finkelstein4-the disturbance in rod function occurs early, isdiffuse, and may be so severe that thresholds inthe dark adapted eye are mostly mediated by cones.

The case of the 12-year-old boy illustrated by

Fig. 9 falls into this category. In a second subtype-type 2-there is a combined elevation in rod and conethresholds. The patient of Fig. 10 is an example of thisgroup. An extensive analysis ofmore than 50 patientswith autosomal dominant RP will be dealt with in a

subsequent paper (Lyness, Ernst, Quinlan, Clover,and Parker, in preparation).While our main use for the perimeter/adaptometer

has been in a study of retinitis pigmentosa, it mayhave important applications in other clinical fields.The interpretation of the dark adaptation curves

shown in Fig. 11 as due to delayed cone recoveryclearly rests on the 2-colour technique.To maximise the effects of the transition from a

scotopic to a photopic spectral sensitivity function theoptimum choice for the 2 test stimuli is a pair of*monochromatic lights with wavelengths of about 480and 670 nm. Unfortunately, until recently, com-mercially available LEDs with a sufficiently high lightoutput have produced their peak emission in therange 560 to 635 nm. In developing the perimeter webegan with a prototype which used Hewlett-PackardLEDs, 5082-4958 (green, dominant wavelength afterfiltering 550 nm) and 5082-4658 (red, dominantwavelength 626 nm). The maximum value for thegreen-red index was about 1P2. We were able toincrease this to 1-6 by exploiting the greater spatialsummation properties of rod compared with conevision. Instead of a single green LED as one of thetest stimuli we arranged 6 to form a ring around thered LED. The light from the ring, when diffused,appeared as an annulus. With this arrangement the

I 6 750 60 70

440



j.com/

Br J O


ownloaded from

http://bjo.bmj.com/

An automated static perimeter/adaptometer using light emitting diodes

readings for green and red log thresholds no longerbelong to the same spectral log threshold function.Instead the former is part of' an 'annulus' functionwhich lies about 0 4 log unit below the 'spot' functionof which the red reading is a part. Analysis ofthresholds measured with spots and annuli at variouseccentricities and also comparisons of patientperformance on the Tubinger and LED perimeterssuggested that no major problems would beencountered with the annulus/spot arrangement, anduseful information from a number of patients wasobtained with the technique. Nonetheless it doesintroduce an additional complication into the inter-pretation of the green-red index. Therefore as soon as

the Stanley LEDs became available-in particularthe ESBR5501 with a light output at 660 nm-thestimuli were changed to 2 single LEDs. The change inwavelength of the red light from 626 to 660 nm

increased the maximum index value from 1 2 to near

2-0 log units. The change in the green light from 550to 530 nm has contributed a further 0 1 log unit. Thisperformance now approaches that of a Tubingerperimeter. In future it may be possible to increase themaximum value of the index further through the use

of a blue LED in place of the green one. Such LEDsare now commercially available.To enable us to make comparisons of data obtained

with our earlier perimeter and the one now in use wehave expressed our results in terms of the extent towhich a patient's performance at a given eccentricitydeviates from normal perfonnance for the same

stimuli. This extent is quantified as a log thresholdelevation. It has therefore been convenient toconstruct an index ofscotopic/photopic function fromthe difference between log threshold elevations forgreen and red stimuli. Since for most regions of thevisual field normal thresholds for both green and redstimuli are characteristic of the scotopic function, thegreen-red index is in general ameasure ofthe patients'deviation from a scotopic threshold match for the 2stimuli. A near zero value indicates that the patient isusing his/her rods for threshold detection, whereas anear maximum value indicates that the patient isdependent on cone function. By contrast, Massof andFinkelstein34 use a different measure to identify rodor cone function, namely, the difference between thevalues given by a patient for the absolute thresholdluminances of red and blue-green lights. Since theluminances are expressed in photopic units, thedifference (approximately) expresses a patient'sdeviation from a photopic threshold match. In fact,for reasons discussed by Massof and Finkelstein4 a

value of about -0 3 rather than zero is expected whenthe patient is using cone function for thresholddetection and a value of 1-9 when he/she is using rodfunction. However, the range of the Massof and

Finkelstein measure (2.2 log units) correspondsapproximately to the maximum value of the index,and the 2 measures can be compared.We have not monitored eye position with the

present instrument. While it would be possible tomodify it to do so by the use of an infrared televisioncamera or a commercially available eye movementmonitor, the cost would be high, and such a changewould undoubtedly introduce a new set of problems(reduction of light, restriction of view, etc.). Ingeneral, patients coming to assist in research are highlymotivated, and the importance of good fixation canbe impressed upon them. Moreover, most patientswith RP have reasonable central vision and are able toachieve good fixation, as judged from observationson the Goldman perimeter. The results shown inFigs. 7 and 8 provide examples that in practicepatients can achieve satisfactory fixation.The great advantages of our instrument lie in the

ease with which it can be used even by a relativelyunskilled operator, the fact that it does not requirethe operator's full-time attention throughout thewhole test, and the fact that the results are graphicallydisplayed even during the test, while a full analysis isavailable immediately on completion. This has meantthat it has become practical for us to include 2-colourstatic perimetry as a routine part of a clinicalinvestigation of a large number of patients rather thanas an exceptional test to be carried out laboriouslyonly on a selected few patients.

Our thanks are due to Miss C. Wheeler for making the measurementson the Tubinger perimeter and to Mr R. J. H. Smith for allowing us touse his instrument, to the workshop of the Institute of Ophthal-mology for help in the construction and maintenance of the LEDperimeter, especially to Mr G. Mould, and to Mr G. Joseph, Mr R.Carter, and Mrs K. Shah for technical assistance. We gratefullyacknowledge financial support from the UK Medical ResearchCouncil, the British Retinitis Pigmentosa Society, the AmericanRetinitis Pigmentosa Foundation, and the Golden Nugget Fund.Throughout the project we received advice and encouragement fromProfessor Alan Bird and Mr Barrie Jay. We would especially like tothank all those persons who volunteered to act as normal observers inthe calibration of the instrument.

References

I Harms H. Die Bedeutung einer einheitlichen Prufereise allerSehfunktionen. Ber Dtsch OphthalmolGes 1960; 63: 281.

2 Massof RW, Johnson MA, Finkelstein D. Peripheral absolutethreshold spectral sensitivity in retinitis pigmentosa. Br JOphthalmol 1981; 65:112-21.

3 Massof RW, Finkelstein D. Rod sensitivity relative to conesensitivity in retinitis pigmentosa. Invest Ophthalmol Visual Sci1979; 18:263-72.

4 Massof RW, Finkelstein D. Two forns of autosomal dominantprimary retinitis pigmentosa. Doc Ophthalmol 1981; 51: 289-346.

5 Duke-Elder S, Smith RJH. In: Duke-Elder S, ed. System ofophthalmology. London: Kimpton, 1962: 7: 399-400.

441



j.com/

Br J O


ownloaded from

http://bjo.bmj.com/

442 W. Emst, D. J. Faulkner, C. R. Hogg, D. J. Powell, G. B. Arden, and Vaegan

6 Gage S, Evans D, Hodapp M, Sorensen H. Hewlett-Packard quantity: principles of measurement. Paris: Bureau central de laoptoelectronics application manual. New York: McGraw-Hill, CIE, 1978.1977. 8 Zeavin BH, Wald G. Rod and cone vision in retinitis pigmentosa.

7 CIE Technical Committee (Vision) 1.4. Light as a true visual AmJOphthalmol 1956; 42: 253-69.



j.com/

Br J O


ownloaded from

http://bjo.bmj.com/

automated static perimeter/adaptometer light emitting diodes · in this paper we describe a static...

Documents