Scanning Laser Entoptic Perimetry for the Detection of Age-RelatedMacular Degeneration

Objective  To determine the sensitivity and specificity of scanning laser entopticperimetry for detecting visual function damage due to age-related maculardegeneration (ARMD).

Methods  We measured the presence or absence of visual field disturbances byentoptic perimetry and determined the severity of ARMD based on masked readingsof fundus photographs. A prospective masked study comparing the findings ofentoptic perimetry with fundus photographs was performed. We recruited 91patients with ARMD and 24 patients without ARMD during ophthalmologic visits.An appropriate institutional review board approval was obtained for the project.The main outcome measure was the detection of visual scotomata.

Results  Scanning laser entoptic perimetry had an overall sensitivity of 82%and a specificity of 100% for the detection of ARMD. The sensitivity for earlystages of the disease is greater than 70%, and increases to above 90% formoderate to late stages.

Conclusion  Scanning laser entoptic perimetry is a specific and sensitive test fordetecting ARMD, even at the earliest stages when patients are typically asymptomatic.

Age-related macular degeneration (ARMD) is the leading cause of severeirreversible visual loss in elderly Americans.¹ Thenumber of cases of ARMD in the United States has been predicted to increasefrom 2.7 million in 1970 to 7.5 million by 2030.² Theprevalence of ARMD also increases dramatically with advancing age. The overallprevalence of any type of ARMD is approximately 20% in the 65- to 74-year-oldage group and up to 35% in the 75- to 84-year-old population.^3,4 ExudativeARMD has a 3-year incidence of approximately 1% in Americans after the ageof 65 years.^4,5 High-risk drusenare present in 13% of the elderly population.⁶

Ophthalmoscopic findings of patients with ARMD include 1 or more ofthe following: the presence of drusen (yellow deposits below the retinal pigmentepithelium [RPE]), hyperpigmentary and hypopigmentary changes of the RPE,atrophic macular degeneration (a well-defined area of atrophy, called geographicatrophy, or another atrophy of the RPE and choriocapillaris), and neovascularmacular degeneration (choroidal neovascularization [CNV], serous or hemorrhagicdetachment of the RPE, and subsequent scarring of the macular area).³ Most ARMD patients have the dry type (including drusen,RPE changes, or RPE atrophy). Of the patients, 10% develop the wet type ofARMD (including CNV, RPE detachment, RPE tears, fibrovascular disciform scarring,and vitreous hemorrhaging).^3,7

The identification of patients at risk for developing the exudativeform of ARMD allows for them to be monitored for early symptoms of CNV. Furthermore,in these at-risk patients, prophylactic treatment, such as vitamins⁸ or laser photocoagulation of soft drusen, may be feasible.⁹ Patients with only drusen in the early form of ARMDmay be asymptomatic but are at higher risk of developing CNV than the healthypopulation.^4-6

Our goal was to determine if entoptic perimetry could detect such centraldisturbances in vision despite good Snellen visual acuity in patients withearly ARMD. We developed a scanning laser device to stimulate the retina,and used it as an en-toptic perimeter to test for visual disturbances. Thetechnique of entoptic perimetry was first described in 1989 using random visualnoise patterns.^10-13 Repressionof extrafoveal scotomata is due to the Troxler effect.¹² Inbrief, a fixed spot of light provided to the peripheral visual field willdisappear from view if saccades are suppressed or if the light remains onthe same area of the retina. We, therefore, used a novel scanning laser–basedentoptic perimeter, which screened the central 60° of vision (30°radius) in patients with and without ARMD.

The Amsler grid is a perimetric tool by which patients look at a 10°× 10° grid of line spaces at 1° intervals, and is designed tomeasure central visual function. While easy to administer and interpret, theAmsler grid is relatively insensitive to relative scotomata that indicateearly stages of disease. Other techniques to detect ARMD include frequency-doublingperimetry, but the researchers conclude that commercial units would requiremodification to detect the small relative scotomata present in patients withearly ARMD.¹⁴ Loewenstein and associates studiedthe use of a hyperactivity computerized visual field test, and found bettersensitivity than that of the Amsler grid in detecting ARMD; however, the testis not widely used.¹⁵

Indeed, the Amsler grid is not widely used by nonophthalmologists. Otherpsychophysical tests to screen for ARMD have been evaluated, including noisefield perimetry¹⁶ and multifocal electroretinography.^17,18 Test results of shape discrimination¹⁹ and contrast sensitivity²⁰ arealso abnormal in ARMD patients and may have a role in screening as well.

We have developed the technique of entoptic perimetry, initially basedon the work of Aulhorn and Kost,¹³ Ramachandran,¹⁰ and Ramachandran and Gregory.¹¹ Thisnoninvasive test requires a brief instructional period. Testing time is relativelyrapid (usually <1 minute), and the test can be potentially self-administered.Our early versions of the test used a television screen or a computer monitorto screen for cytomegalovirus retinitis. In subsequent studies, patients observedthe stimulus in a scanning laser device. We found that patients were ableto detect visual field abnormalities due to various retinal diseases (cytomegalovirusretinitis, diabetic retinopathy, and macular degeneration) and damage dueto glaucoma within the central 120° of the visual field.^21,22

During entoptic perimetry, a patient is shown a monochromatic fieldof random particle motion. While looking at the central fixation spot, patientswith dense scotomata (due to cytomegalovirus retinitis) reported that in someareas of the visual field the random particle motion disappears and is replacedwith gray. If a patient has relative scotoma (due to early glaucoma, cotton-woolspots, or another retinal injury), the patient typically reports that thereis some residual particle motion in the affected areas; these areas are qualitativelydifferent from the remainder of the visual field. The particle motion of entopticperimetry seems to circumvent the Troxler effect.^21-23

We recruited patients who were diagnosed as having ARMD in 1 or botheyes. We also recruited a control group of volunteers who were classifiedas healthy in the same clinic. Patients were recruited by one of us (W.R.F.)during ophthalmologic visits for the treatment of ARMD from the retina clinicsat the Shiley Eye Center, University of California, San Diego. There was nominimum requirement for visual acuity. Participation of all subjects was voluntary.We received informed consent from each patient and healthy control beforetesting.

The stimulus was delivered through a virtual retinal display (VRD) developedby Microvision, Inc, Bothell, Wash. The VRD delivered a monochromatic imagevia a scanning laser beam projected onto the retina from the standard videographicsarray output signal generated by a portable computer (Macintosh G3 Powerbook).Each pixel in the image could have 1 of 2 values: off (black) or on (deepred [635 nm]). The stimulus was previously described in detail.²¹ Inbrief, at a given time, each pixel was randomly assigned either the on oroff value. These values changed over time. The overall effect was that thevisual field was filled with random particle motion. There were no discerniblepatterns within the image over time.

The appearance of the VRD is similar to a telephotographic camera lens.Patients viewed the image through a 1-mm aperture at one end of the VRD. Theopposite end was connected to the video input from a computer. The VRD wassecured to the table to reduce vibrations. Patients minimized head movementsby placing their chin and forehead against an adapted slitlamp headrest thatwas also secured to the table. Previous studies^21,22 provedthat minimizing head movement increased the sensitivity of the test. Thisalso allowed people to fixate more readily on the narrow exit aperture.

The computer generating the stimulus passed the video signal througha splitter. One cable from the splitter was attached to the VRD, while thesecond was attached to a separate computer monitor that displayed the imagein black and white. The monitor allowed the technician to control the stimulusand view the program without requiring the patient to move from the VRD.

Before testing, each patient underwent a dilated ophthalmoscopic examinationusing an indirect ophthalmoscope and slitlamp biomicroscopy followed by fundusphotography.

The technician briefed each patient on the purpose of entoptic perimetryand the testing requirements. Patients who agreed to the testing procedurereceived a brief overview of the procedure, including potential risks, andthen were given a consent form. Following informed consent, patients receiveda tutorial on en-toptic testing. The instructional component rarely took morethan 2 minutes. By using the computer monitor, the experimenter showed thestimulus and the fixation point to the patient. Patients were told that ifthere were areas where the particle motion changed or looked qualitativelydifferent, they should use the digital pen to outline the edges of the areaof disturbance. The digital pen allowed the user to draw virtual lines usinga touch tablet and stylus on the desk next to the stimulus. The lines appearedon the computer and to control the stimulus (described later). When satisfiedwith a drawing, it would be saved in the computer. If the patient believedthat the drawing was inaccurate, the patient would be allowed to redraw it.This test was performed on both eyes unless there was a confounding variable,such as no light perception in the fellow eye.

The entoptic program had 2 display modes: stimulus and drawing. Whenin the stimulus mode, random particle motion filled the screen, except forthe central fixation cross. Patients switched from stimulus to drawing modeby bringing the digital pen close to the tablet (within approximately 1 cm).The screen then became uniformly colored (white on the computer monitor andred on the VRD). Moving the pen just above the tablet changed the locationof the cursor. This allowed the patient to relocate the pen to areas of visualdisturbance without drawing a line. Conversely, pressing the pen against thepad and moving it slowly with a firm but light pressure drew a line on thescreen. If the patient made a mistake or was unsatisfied with any aspect ofa drawing, the experimenter either used the undo function to remove the previousline or cleared the drawing pad and allowed the patient to redo the trace.

One possible confound in the stimulus is that the perception of thevisual field disturbance disappears when switching from stimulus to drawingmode. This was mitigated by allowing the patient to switch between modes simplyby varying the distance of the digital pen from the tablet. Patients receivedinstructions to use a trial-and-error method by which they would flip betweenstimulus and drawing mode to carefully trace out edges of all visual fielddisturbances. We allowed patients to practice drawing and switching betweenmodes throughout the instructional phase of testing. Each patient was instructedto outline, as best as the patient could, the edge of the extent of any areaof the stimulus that differed qualitatively from the monochromatic particlemotion.

During the testing phase, one eye was patched. The patient was alignedto the VRD. The subject’s head was placed in the headrest, and the VRDwas moved into position by the technician until patients reported that theywere able to see the entire screen, filled with particle motion, as they hadobserved on the computer monitor. No patient reported any problems in alignmentor ability to view the stimulus within the VRD once properly aligned. Thenarrow aperture proved to be a benefit because it forced the patient to focuson the stimulus at all times. If the patient’s gaze or attention wandered,the patient would not be able to perceive the stimulus. Patients were toldto concentrate their attention on the central cross. The stimulus was started,and patients were then allowed to use the pen to switch between modes andrecord the locations of disturbance. The field of view was 30°.

As previously noted, fundus photographs of each patient were taken aspart of the normal examination during the visit. We determined the presenceof retinal damage due to ARMD by indirect ophthalmoscopy and confirmed byfundus photographs. Scoring of the photographs was performed by 2 ophthalmologists(W.R.F. and M.E.-B.), and reviewed by an expert in retinal disease (W.R.F.).The ophthalmologists classified each eye into 1 of 8 categories by gradingstereoscopic color photographs and angiograms. The following categories wereused: mild drusen, moderate drusen (many soft drusen), mild geographic atrophy,moderate to advanced geographic atrophy, pigment epithelial detachment, untreatedCNV, inactive previously treated CNV, and untreated but regressed CNV. Weresolved any conflicts in categorization by consensus.

An expert psychophysicist (D.J.P.) scored all entoptic tracings forthe presence or absence of a visual field disturbance. A given eye was classifiedas having no visual field disturbance if the patient made no marks on thescreen with a digital pen (negative), while having any marks within the visualfield was classified as a disturbance in visual function (positive). The psychophysicistwas masked to ophthalmologic findings until completion of all scoring.

For each classification, we determined the ratio of eyes that were positiveby entoptic perimetry and fundus photography (true positives by entoptic perimetry)to the eyes positive by fundus photography (the gold standard). Similarly,specificity was determined as the ratio of the number of eyes scored negativeby entoptic perimetry, which were also negative by fundus photography (truenegatives by entoptic perimetry), to the number of eyes scored negative byfundus photography (the gold standard).

Sensitivities were calculated for each of the 8 classifications of ARMDand the combination of those 3 stages that compose early ARMD. Specificitywas derived from healthy control eyes.

We recruited 91 patients (41 women and 50 men) with ARMD, for a totalof 171 eyes (mean ± SD age, 68.76 ± 10.29years). We did not test 11 fellow eyes from this group because of poor centralvision, resulting in the inability to perceive the stimulus. We recruited24 healthy control patients, for a total of 43 eyes (mean ± SDage, 64.30 ± 15.81 years). We did not test 5 fellow eyesfrom these patients. There were no significant differences in mean age betweengroups (P>.40).

The Table provides a summary ofresults, including stratification by severity of disease. Overall, we foundthat scanning laser entoptic perimetry has a mean ± SD sensitivityof 0.82 ± 0.04 and a specificity of 1.00.

The sensitivity for early stages of the disease (mild and moderate drusenand mild geographic atrophy, with a visual acuity better than 20/40), whenpatients typically remain asymptomatic, ranges from 0.70 to 0.93 (mean ± SD,0.77 ± 0.06). The sensitivity of entoptic perimetry increasesfor later stages of the disease, averaging approximately 0.93. Patients areable to detect visual field disturbance due to CNV with a high sensitivity(treated, 0.92; and untreated, 1.00). We had 4 special cases of advanced ARMDin which diagnoses crossed more than 1 grouping boundary. For all 4 eyes,patients reported entoptic disturbances.

It was previously demonstrated that scanning laser–based entopticperimetry is an excellent screening tool for detecting retinal injury dueto various diseases, including ARMD.²¹ In thatstudy, relatively few ARMD patients were tested and many had moderate to advanceddisease, potentially limiting the interpretation of the results. This studydirectly addresses the limitations of the previous studies. We tested manyeyes (171 vs 4) and developed a method of stratification by disease stage.This provides a metric for determining how sensitive entoptic perimetry (orany screening test) is for detecting ARMD during early vs late stages of thedisease.

The results for later stages of the disease are consistent with previousfindings.^21-23 Inthis study, we found that the overall specificity (mean ± SD,1.00 ± 0.00) was high, but similar to what was previouslyreported.^21-23 Wealso found that the overall sensitivity of entoptic perimetry was 82%. However,this may be an underestimation of the true sensitivity of entoptic perimetry.Our patient cohort is biased toward patients with early forms of the disease.More than half of the eyes in this study (56.1%) were classified within the2 mildest forms of ARMD. These forms have the shallowest scotomata and thelowest sensitivity. We recruited many of these patients because all ARMD patientspass through this stage and these are the patients who would benefit mostfrom emerging therapies.

For early ARMD, our a priori assumption was that en-toptic perimetrywould have a relatively low sensitivity for detecting these types of scotomata.In contrast, the results for the sensitivity of entoptic perimetry for detectingearly and moderately advanced stages of ARMD were impressive. We found thatfor patients classified as having drusen we could detect visual function losswith a sensitivity greater than 70%, and for those with mild geographic atrophyat a rate of 93%. The overall sensitivity for detecting any type of ARMD,capturing patients who are typically asymptomatic, was 82%. These resultssuggest that, given an optimized stimulus, we can detect three quarters ofall early forms of ARMD. En-toptic perimetry depends on responses from everylevel of the visual system, from the photoreceptors to the brain, and abnormalitiesthat are detected cannot necessarily be located to one specific area. Thisis important for primary care physicians or optometrists serving geriatricpopulations. If a physician can detect ARMD early and refer to an ophthalmologist,it is possible that future therapies will reduce the total amount of visionlost. Standard screening tests like the Amsler grid and Humphrey Field Analyzercannot always detect visual function loss due to ARMD at this early stageof the disease. We did not, however, compare scanning laser entoptic perimetrywith standard automated perimetry directly. The latter is more expensive andtime-consuming.

These results suggest that scanning laser entoptic perimetry is an effectivescreening test for ARMD. Entoptic perimetry has another advantage in thatour entoptic perimetry setup is portable. This test could be easily implementedby primary care physicians and transported to outpatient clinics and underservedcommunities. Because ARMD is a disease primarily affecting geriatric populations,house staff could provide periodic screenings at rest homes. The support staffat any of these locations could administer this test with a minimum time expenditure.This will allow asymptomatic patients with ARMD to be referred to ophthalmologistsbefore severe damage occurs to central vision.

The VRD has 3 distinct advantages over flat-screen technologies whenimplementing entoptic perimetry. It has the capability to display a stimulusout to 30° eccentric from the fovea when fixated centrally. Because ARMDis a disease primarily of the central retina, there was no reason to screenoutside these areas, as done in other studies. The VRD also produces a relativelyhigh signal-noise ratio relative to computer monitors. Images from the VRDare projected directly into the eye at virtual infinity. This is an immensebenefit when testing geriatric populations that may have latent cataracts,media opacities, and high refractive errors. The VRD, therefore, displaysimages on the retina more independent of the visual system’s optics,reducing the importance of optical aberrations of the anterior segment andeliminating the need to perform exact refractions for each patient. We didnot analyze the correlation between location of retinal abnormality and entopticdefects. Prior work^21-23 hasshown that there is a close correlation between these 2 factors. In ARMD patients,however, abnormalities are diffuse and may be difficult to locate with precision.

Entoptic perimetry is not intended to replace the need for ophthalmologistsor other screening tools, like visual field perimetry. This screening testwill best serve primary care physicians and optometrists in detecting ARMDin its early stages.

Correspondence: William R. Freeman, MD,Department of Ophthalmology, Shiley Eye Center, University of California,San Diego, 9415 Campus Point Dr, La Jolla, CA 92093-0946 (freeman@eyecenter.ucsd.edu).

Submitted for Publication: August 6, 2002;final revision March 1, 2004; accepted April 22, 2004.

Financial Disclosure: None.

Funding/Support: This study was supported bygrants E407366 (Dr Freeman) and EY11961 (Dr Plummer) from the National EyeInstitute, Bethesda, Md; and by Research to Prevent Blindness, New York, NY.