Optic nerve and retinal

Optic nerve and retinal nerve fiber layer analyzers

in glaucoma

David S. Greenfield, MD

There is mounting evidence that retinal nerve fiber layer(RNFL) loss precedes detectable visual field loss in earlyglaucomatous optic neuropathy. However, examination andphotography of the RNFL is a difficult technique in manypatients, particularly older individuals, and eyes with smallpupils and media opacities. It is subjective, qualitative, variablyreproducible, and often unreliable. Furthermore, optic nervehead and RNFL photography is time consuming, operatordependent, has limited sensitivity and specificity, and requiresstorage space. Imaging technologies have emerged whichenable clinicians to perform accurate, objective, andquantitative measurements of the RNFL and optic nerve headtopography. There is good agreement between suchmeasurements and clinical estimates of optic nerve headstructure and visual function. The reproducibility of theseinstruments suggests that they have the potential to detectstructural change over time. This report will review thetechnological principles, reproducibility, sensitivity andspecificity, capacity to detect glaucomatous progression,and limitations of currently available ocular imagingtechnologies. Curr Opin Ophthalmol 2002, 13:68–76 © 2002 Lippincott

Williams & Wilkins, Inc.

Glaucoma is an optic neuropathy characterized by a typi-

cal pattern of visual field loss and optic nerve damage

resulting from retinal ganglion cell death caused by a

number of different disorders that affect the eye. Most,

but not all, of these disorders are associated with el-

evated intraocular pressure (IOP), which is the most im-

portant risk factor for glaucomatous damage. Although

clinical examination of the optic nerve head has been

considered to be the most sensitive test for detecting

glaucomatous damage, evidence suggests that examina-

tion of the retinal nerve fiber layer (RNFL) may provide

important diagnostic information [1-4]. Accurate and ob-

jective methods of detecting disc and RNFL abnormali-

ties, and their progression, would facilitate the diagnosis

and monitoring of glaucomatous optic neuropathy.

Clinical examination and photography of the RNFL is a

difficult technique in many patients, particularly older

individuals, those with small pupils, and subjects with

media opacities. It is subjective, qualitative, variably re-

producible, and often unreliable. In addition, optic nerve

head and RNFL photography is time consuming, opera-

tor dependent, has limited sensitivity and specificity,

and requires storage space. Recently, new technologies

have emerged which enable clinicians to perform accu-

rate, reproducible, objective, and quantitative measure-

ments of the retinal nerve fiber layer and optic nerve

head topography.

Confocal scanning laser ophthalmoscopy (CSLO), a

technology embodied in the Heidelberg Retinal

Tomograph (HRT, Heidelberg Engineering, Heidel-

berg, Germany), enables the operator to evaluate three-

dimensional characteristics of optic nerve head topogra-

phy quantitatively [5-8]. Thirty-two coronal sections of

the optic nerve head are acquired over a depth of ap-

proximately 3.5 millimeters, and a color-coded topo-

graphic map of the optic nerve head is generated.

Scanning laser polarimetry (SLP) is a technology embod-

ied in the GDx Nerve Fiber Analyzer (Laser Diagnostic

Technologies, Inc., San Diego, CA) employs a confocal

scanning laser ophthalmoscope and an integrated polar-

imeter. It evaluates the thickness of the RNFL by uti-

lizing the birefringent properties of the retinal ganglion

cell axons [9,10]. As polarized light passes through the

RNFL and is reflected back from the deeper layer, it

undergoes a phase shift. The change in polarization, re-

The Department of Ophthalmology, The University of Miami School of Medicine,Bascom Palmer Eye Institute, Miami, Florida, USA.

Correspondence to David S. Greenfield, MD, Bascom Palmer Eye Institute, 7108Fairway Drive, Suite 340, Palm Beach Gardens, FL, 33418; e-mail:[email protected]

Current Opinion in Ophthalmology 2002, 13:68–76

Abbreviations

CSLO confocal scanning laser ophthalmoscopyHRT Heidelberg Retinal TomographIOP intraocular pressureOCT optical coherence tomographyRNFL retinal nerve fiber layerSLP scanning laser polarimetry

ISSN 1040–8738 © 2002 Lippincott Williams & Wilkins, Inc.

68

ferred to as retardation, is proportional to the thickness of

the birefringent medium, and is measured to give an

index of RNFL thickness.

Optical coherence tomography (OCT, Zeiss-Humphrey

Systems, Dublin, CA) is a noninvasive, noncontact,

transpupillary imaging technology that can image retinal

structures in vivo with a resolution of 10 to 17 microns

[11,12]. Cross-sectional images of the retina are produced

using the optical backscattering of light in a fashion

analogous to B-scan ultrasonography. The anatomic lay-

ers within the retina can be differentiated and retinal

thickness can be measured [13].

This report will review practical applications and prin-

ciples underlying these posterior segment-imaging tech-

nologies with emphasis upon strengths and limitations of

each technology.

Confocal scanning laser ophthalmoscopyTechnological principles

Confocal scanning laser ophthalmoscopy employs a 670 nm

diode laser beam as a light source and scans the retina in

x- and y- directions [14,15]. Light originating from the

illuminated area passes through a diaphragm (pinhole) in

a plane optically conjugate to the retina. Planes unfo-

cused at the aperture are blocked by the diaphragm and

do not reach the detector. Each image contains 256 x 256

pixels (picture-elements); each pixel represents the reti-

nal height at that location relative to the focal plane of

the eye. Image acquisition and processing takes approxi-

mately 1.6 seconds. Thirty-two coronal sections are ob-

tained progressing from anterior to the optic nerve head

through the retrolaminar portion of the nerve head. The

axial distance between two adjacent sections is 50 to

75 m generating an axial range of 1.5 to 3.5 mm.

A standard reference plane is established parallel to the

peripapillary retinal surface and is located 50 microns

posterior to the retinal surface along a circle concentric

with the optic disc margin in a temporal segment be-

tween 350° and 356°. Neural rim is defined as tissue

within the optic disc margin and above the reference

plane. Optic cup is defined as tissue within the disc

margin and below the reference plane.

The optic disc margin is outlined and a color-coded

depth map is created from a mean topographic image

using a software algorithm (Fig. 1). Stereometric param-

eters of optic nerve head topography are generated rela-

tive to the reference plane including rim area and vol-

ume, cup area and volume, cup-disc area ratio, mean

retinal nerve fiber layer thickness, and retinal nerve fiber

cross-sectional area. Parameters independent of the ref-

erence plane include mean and maximum cup depth,

height variation contour, and cup-shape measure. A nor-

mal retinal height variation diagram demonstrates a

“double-hump” pattern corresponding to the thicker

retinal ganglion cell axons along the superior and inferior

portions of the optic nerve head.

Reproducibility

Various investigators have reported high levels of repro-

ducibility using this technology [5,15,16] Brigatti et al. [7]

found that topographic variability correlated with the

steepness of the corresponding region. Greater variabil-

ity was found at the edge of the optic disc cup and along

blood vessels. Weinreb et al. [14] have determined that

measurement reproducibility is improved from 35.5 µm

to 25.7 µm when a series of three examinations are ob-

tained instead of a single image analysis. Based upon

these data, acquiring three images per eye and creation

of a mean topographic image is recommended. Finally,

Zangwill et al. [17] have shown that image reproducibil-

ity is improved with pupillary dilation, particularly in

eyes with small pupils and cataract.

Sensitivity and specificity

Various investigators have reported topographic differ-

ences between normal, ocular hypertensive, and glauco-

matous eyes. It is essential to emphasize that the char-

acteristics of the study population will influence the

discriminating power involved in differentiating glauco-

matous from nonglaucomatous eyes. Determination of

sensitivity and specificity parameters is fundamentally

linked to the severity of glaucomatous damage among

the cohort studied. For any given technology, an instru-

ment will appear to be more sensitive if it is used to

separate eyes with advanced glaucoma from normal sub-

jects compared with eyes with mild glaucoma.

Heidelberg Retinal Tomograph employs software with

various statistical analyses to discriminate normal from

Figure 1. Confocal scanning laser ophthalmoscopy

topographic map

A patient with moderate normal-tension glaucoma shows loss of the inferiorneuroretinal rim (green) and associated stereometric parameters. There is a focaldepression in the double-hump pattern of the height variation diagramcorresponding to the decreased inferotemporal quadrant height (below).

Optic nerve and retinal nerve fiber layer analyzers in glaucoma Greenfield 69

abnormal optic discs. These include a multivariate dis-

criminant analysis based upon rim volume, height varia-

tion contour, and cup shape measure adjusted by age

[18], ranked-segment distribution curves [19,20], and re-

gression analysis using a normative database of 80 normal

eyes from 80 white subjects with a mean age of 57 years

[21]. The confidence interval limits derived from the

later are used commercially to generate the Moorfield’s

Regression Classification Score (normal, borderline, or

outside normal limits). Wollstein et al. [21] reported a

84.3% sensitivity and a 96.3% specificity for separating

normal and early glaucomatous eyes by taking into ac-

count the relation between optic disc size and the rim

area or cup-to-disc area ratio. In a different study, Woll-

stein et al. [22•] determined that by taking into account

the optic disc size, HRT image analysis was superior in

sensitivity (84.3%) for detection of early glaucoma com-

pared with expert assessment of stereoscopic optic disc

photographs (70.6%).

The sensitivity and specificity of various HRT param-

eters has been investigated and varies widely ranging

from 62% to 94% and 74% to 96%, respectively [18,23–

27]. Wide variability in discriminating power may be ex-

plained in part by variable sample size, definitions of

glaucoma, and varying degrees of glaucomatous optic

nerve damage. A recent study by Miglior et al. [28•]

found fair to poor agreement (� statistic 0.28-0.48) be-

tween visual field examinations and HRT classifications

among a population of 359 eyes (55 normal, 209 with

OHT, and 95 with moderate POAG, average visual field

mean defect –7.6 dB) The sensitivity and specificity of

the HRT examination were, respectively, 80% and 65%,

using the Mikelberg multivariate discriminant analysis

[18], and 31 to 53% and 90 to 92%, using ranked-segment

distribution curve analysis [19,20].

Using various HRT summary data including the reflec-

tance image, double-hump graph, stereometric analyses,

and HRT classification using a multivariate discriminant

function [18] and ranked segment analysis [19,20],

Sanchez-Galeana [29•] evaluated the sensitivity and

specificity for discriminating between 50 normal eyes

and 39 eyes with early to moderate glaucoma (average

visual field mean defect –5 dB). Masked observers were

used to generate an HRT classification (normal, glauco-

matous, or undetermined) and similar classifications

were generated using other imaging technologies (see

below). Using these summary data collectively, investi-

gators reported a sensitivity and specificity for the HRT

ranging from 64 to 75% and 68 to 80%, respectively.

Detection of progression

Essential elements for change detection algorithms have

been previously reviewed [30]. An accepted gold stan-

dard must exist for establishing change. Surrogate mea-

surement parameters are necessary with little biological

variability and relevance in the course of the disease.

High instrument reproducibility is essential with known

limits of variability in normals and persons with disease.

Statistical criteria must be established for differentiating

biological change from test-retest variability. Finally,

multicenter prospective validation must be established

with comparisons against an accepted gold standard.

Confocal scanning laser ophthalmoscopy strategies for

change detection exist including serial analyses of global

and regional topographic indices (eg, cup-disc ratio, cup

volume, and cup-shape measure), and color-coded

(red/green) significance indicators of change relative to

baseline. Chauhan et al. [31••] have described a sophis-

ticated change analysis algorithm based upon a probabi-

listic approach using variability estimates that employs

clusters of 4 x 4 pixels to create superpixels. Three fol-

low-up images are compared with a baseline image, and

a change-probability map is created, characterized by ar-

eas with significant progression illustrated in red.

Strengths of this algorithm include the potential ability

to differentiate biological change from test-retest vari-

ability, however it has not been validated in prospective

clinical trials. Moreover, topographic measurements are

dependent upon intraocular pressure and postoperative

and diurnal changes in IOP have been reported to pro-

duce changes in optic disc topography thereby confound-

ing detection of progression.

Two reports have described HRT detection of change.

Chauhan et al. [31••] described significant topographic

change in one patient with progressive glaucomatous op-

tic disc cupping. Kamal et al. [32] reported topographic

disc changes in a cohort of thirteen ocular hyperten-

sive subjects converting to glaucoma before confirmed

visual changes. This study was limited, however, by

small sample size, reviewers unmasked to diagnosis,

absence of a control arm of OHT non-converters, and

inability to differentiate biological change from test-

retest variability.

LimitationsTechnological limitations exist which limit the discrimi-

nating power for disease detection. The use of a standard

reference plane and need for correct placement of the

disc margin by the operator can influence many of the

topographic outcome variables generated. Moreover,

considerable variability in optic disc morphology exists

among normal eyes. As currently configured, software

algorithms designed to classify subjects as normal or

glaucomatous are based upon dedicated normative data

of approximately 100 eyes which is insufficient for popu-

lation based screening. A uniform consensus regarding

the most appropriate summary measures remains to

be established.

There is evidence that disc topography is dependent

upon intraocular pressure [33] and cardiac pulsation [34].

70 Glaucoma

Postoperative [35,36 ] and diurnal [37] changes in IOP

may produce changes in optic disc topography thereby

confounding detection of glaucomatous progression. In

addition, CSLO cannot discern vessel shift or other non-

quantitative features (eg, pallor or disc hemorrhage)

often associated with progression. Finally, as with perim-

etry, short and long-term fluctuation exists and confi-

dence intervals need to be validated to interpret mea-

surements obtained.

Scanning laser polarimetryTechnological principles

Scanning laser polarimetry (SLP) is a technology that

provides quantitative assessment of the peripapillary

RNFL using a polarized diode laser light source (780

nm). The parallel arrangement of neurotubules within

the RNFL produces linear birefringence. Thus, changes

in the polarization state may be measured when light

passes through such tissue [9,10,38-40]. The change in

polarization of the scanning beam (retardation) is linearly

correlated to the thickness of the polarizing medium, and

is computed to give an index of RNFL thickness. A

polarization detection unit measures the retardation of

light emerging from the eye; 256 by 256 pixels (65,536)

are acquired in 0.7 seconds and a computer algorithm

calculates retardation at each retinal position.

An anterior segment compensator is incorporated within

the technology to neutralize the polarization effects of

the cornea and crystalline lens. It consists of a fixed re-

tarder to adjust for the corneal retardation and assumes

all individuals have a slow axis of corneal birefringence

15 degrees nasally downward and a magnitude of 60 nm

[41,42 ]. Recent studies have demonstrated that the mag-

nitude [43] and axis [44] of corneal polarization are vari-

able, and are strongly correlated with RNFL thickness

assessments obtained with SLP.

At least three images are acquired using a field of view of

15 x 15 degrees and a baseline retardation map is created.

Images may be obtained through an undilated pupil with

a minimum diameter of 2 mm. However, uniformity in

pupil size is essential when longitudinally evaluating

RNFL measurements.[45] The probability of obtaining

a satisfactory baseline image (mean pixel SD </= 8 µm)

improves from 62 to 98% if the number of scans available

for selection is increased from three to five.[40] The

retardation map represents a false color image with areas

of high retardation displayed in yellow and white, and

areas of low retardation displayed in blue (Fig. 2).

The operator outlines the optic disc margin, and a ten-

pixel-wide measurement ellipse is automatically gener-

ated, 1.75x greater than the disc diameter. A computer

algorithm automatically generates retardation mea-

surements throughout the peripapillary region and along

the measurement ellipse. Average quadrantic measure-

ments, measurement ratios (eg, superior/nasal,

superior/temporal), symmetry measurements between

superior and inferior quadrants, and modulation param-

eters (an indication of the difference between the thick-

est and thinnest parts of the RNFL) are generated. A

neural network number is also calculated which is

thought to reflect the likelihood of glaucoma on a scale of

0 to 100.

Reproducibility

Intraoperator measurement reproducibility has been

shown by Weinreb et al. [10] (mean coefficient of varia-

tion (CV) of 4.5%) and Chi et al. [46] (CV ranging from

3.59–10.20% for both normal and glaucomatous sub-

jects). Swanson et al. [47] found significant interoperator

variability with the NFA I, among 4 operators all of

whom only scanned each of the 11 subjects twice. The

primary source of error was attributed to the variability in

the criterion used for establishing intensity setting. This

problem was subsequently reduced in the NFA II with a

hardware modification to the light system.

Retinal nerve fiber layer thickness measurements using

the NFA II have been reported to have high levels of

measurement reproducibility [40,48]. Hoh et al. [40] de-

scribed excellent intraoperator reproducibility and found

that variability between operators can be minimized by

using a single measurement ellipse acquired from the

original baseline image. As investigators have reported

high levels of measurement variability adjacent to retinal

blood vessels [49,50], an automated blood vessel removal

algorithm has been incorporated in the third generation

device, GDx.

Sensitivity and specificity

As described with CSLO, there is a wide range in RNFL

thickness values among normal individuals and consid-

erable measurement overlap between normal and glau-

comatous eyes may exist. Determination of sensitivity

and specificity parameters is fundamentally linked to the

Figure 2. Scanning laser polarimetry image

A patient with moderate primary open-angle glaucoma shows reducedretardation within the superior arcuate retinal nerve fiber layer bundle. Tworetardation parameters were classified as abnormal (outside 95% confidencelimits, illustrated in red) and four parameters were classified as borderline(outside 90% confidence limits, illustrated in yellow).

Optic nerve and retinal nerve fiber layer analyzers in glaucoma Greenfield 71

severity of glaucomatous damage among the cohort stud-

ied [24]. Sensitivity and specificity values will be greater

in studies involving eyes with advanced glaucoma than

in studies involving eyes with mild to moderate glau-

coma. Tjon-Fo-Song and Lemij [38] evaluated the sen-

sitivity and specificity of the first generation device,

NFA I, for detecting glaucoma among a diverse group of

200 eyes with early to advanced glaucoma (average visual

field mean deviation –10.33 decibels) compared with a

normal population. The sensitivity and specificity was

reported to be 96 and 93%, respectively. Weinreb et al.[51] reported a sensitivity of 74% and specificity of 92%

using a newer version of SLP with a linear discriminant

function to label glaucomatous damage among a popula-

tion with early to moderate glaucoma. Garcia-Sánchez

et al. [52] found the sensitivity and specificity of the GDx

to be 78% and 86%, respectively. The most sensitive and

specific parameters in their study were ellipse modula-

tion, superior/nasal ratio, and maximum modulation.

In a cross-sectional study comparing OCT and SLP, Hoh

et al. [53] found that structural information generated

from both technologies was significantly correlated with

visual function in glaucomatous eyes (average visual field

mean deviation –7.7 decibels). However, retardation pa-

rameters providing summary measures of RNFL thick-

ness (eg, average thickness and integral measurements)

had a weaker correlation with visual field mean defect

(R = 0.17 to 0.27) than with constructed retardation pa-

rameters (eg, modulation scores, ratio parameters, and

number; R = 0.36 to –0.51). Bowd et al. [54] recently

reported that constructed SLP parameters (modulation,

ratio, number, and linear discriminant function values)

have the greatest discriminating power. This is ex-

plained by recent evidence [44] suggesting that interin-

dividual variability in corneal birefringence has falsely

broadened the normative database of RNFL thick-

ness assessments, and reduced the sensitivity and speci-

ficity of this technology. Correction for corneal polariza-

tion axis has been shown to significantly increase

the correlation between RNFL structural damage and

visual function, and significantly improve the discri-

minating power of SLP for detection of mild to moder-

ate glaucoma.

Garcia-Sanchez et al. [29] evaluated the sensitivity and

specificity of the HRT, GDx, and OCT summary data for

detection of early to moderate glaucoma (average visual

field mean defect –5.0 dB) among three masked reviewers

(see Table 1). For the GDx, sensitivity and specificity

ranged from 72 to 82% and 56 to 82%, respectively.

Detection of progression

Scanning laser polarimetry strategies for change detec-

tion exist including evaluation of change in absolute val-

ues of retardation measurements, change in quadrantic

RNFL thickness measurements, change in double-

hump RNFL thickness profile, and color-coded map of

RNFL thickness change relative to baseline. However,

as with OCT, statistical units of change probability are

absent limiting the ability to differentiate change from

measurement variability, and there has been no prospec-

tive validation of this algorithm

Two published reports have described SLP evidence of

change detection in eyes with non-glaucomatous optic

neuropathy. Colen et al. [55] described a patient with

acute nonarteritic anterior ischemic optic neuropathy

who developed progressive loss of retardation over a

5-week period corresponding to a dense altitudinal visual

field depression. Medeiros and Susanna [56] reported

progressive RNFL loss over a 90-day period in a patient

with traumatic optic neuropathy.

Limitations

Employment of a fixed corneal compensator has pro-

duced considerable measurement overlap among normal

and glaucomatous eyes. Variability in corneal polariza-

tion axis (CPA) [57••] and magnitude has been de-

Table 1. Comparison of scanning laser ophthalmoscopy, scanning laser polarimetry, and optical coherence tomography

GDx HRT OCT

Technological principle Birefringence SLO InterferometryPixels 65,000 65,000 50,000Pupillary dilation No No YesReproducibility (CV) 5%–10% [40] 5%–10% [67] 5%–10% [63]Parameters measured Peripapillary RNFL Optic Disc Topography Peripapillary RNFLNormative database 1200 eyes [68] 45, [19] 100 [19] or 112 [21] eyes 150 eyes*Sensitivity [29] 72%–82% 64%–75% 76%–79%Specificity [29] 56%–82% 68%–80% 68%–81%Change detection algorithm Yes Yes YesChange probability algorithm No Yes NoProspective validation of algorithm No No NoEvidence to detect change Yes [55, 56] Yes [31, 32] NoLimitations Fixed corneal compensator;

unable to differentiatevariability from progression

Universal reference plane;topography is dependentupon IOP

Sampling data limited to 100A-scans; unable to differentiatevariability fromprogression

SLO, scanning laser ophthalmoscopy; CV, coefficient of variation.*Personal communication (Zeiss-Humphrey Systems, Dublin, CA).

72 Glaucoma

scribed; there is evidence that CPA strongly effects peri-

papillary retardation measurements (Fig. 3).

Although, there is good one-year stability of CPA mea-

surements [58], long-term stability and the effect of in-

traocular and refractive surgery upon such measurements

remains unknown. Furthermore, anterior and posterior

segment pathology may produce spurious RNFL mea-

surements [59], and caution should be used when inter-

preting images in eyes with ocular surface disease, pre-

vious keratorefractive surgery, media opacification, and

extensive peripapillary atrophy.

Although a change analysis algorithm exists, statistical

units of probability are absent. Thus, biological change

cannot be differentiated from measurement variability.

Finally, prospective studies are necessary to validate

change analysis strategies.

Optical coherence tomographyTechnological principles

Optical coherence tomography (OCT, Zeiss-Humphrey

Systems, Inc., Dublin, CA) is a noninvasive, noncontact,

transpupillary imaging technology which can image reti-

nal structures in vivo with a resolution of 10 to 17 microns

[11,12]. Cross-sectional images of the retina are produced

using the optical backscattering of light in a fashion

analogous to B-scan ultrasonography. The anatomic lay-

ers within the retina can be differentiated and retinal

thickness can be measured [13].

Optical coherence tomography images are obtained us-

ing a transpupillary delivery of low coherence near-

infrared light (850nm) from a super-luminescent diode

laser [11–13,60]. Backscatter from the retina is captured

using the same delivery optics and resolved using a fiber-

optic interferometer set in a standard Michelson con-

figuration. Modulating the reference arm allows longitu-

dinal information to be extracted to the resolution as

defined by the low coherence super-luminescent diode.

Cross-sectional OCT images of the retina are con-

structed from the backscattering information provided

by 100 individual longitudinal A-scans. A digitized,

composite image of the 100 A-scans is produced on a

monitor with a false color scale representing the degree

of light backscattering from tissues at different depths

within the retina.

A minimum pupillary diameter of 5 mm is required to

obtain satisfactory OCT image quality. Images may be

acquired using either a linear or circular scanning beam.

Scanning acquisition time is approximately one second.

A circular scan of the RNFL is generally performed with

a diameter of 3.4 mm (Fig. 4) to avoid areas of peri-

papillary atrophy. Circular scans of this diameter contain

100 axial scans spaced 110 microns apart. This scan is

then converted into a radial image by an automated

“smoothing” technique. A computer algorithm identifies

and demarcates the signal corresponding to the RNFL,

and mean quadrantic and individual clock hours of

RNFL thickness measurements are calculated.

Reproducibility

Schuman et al. [61] evaluated the reproducibility of reti-

nal and RNFL thickness measurements using circular

scans around the optic nerve head in normal and glau-

comatous eyes. Scan diameters of 2.9, 3.4, and 4.5 mm

were evaluated and internal fixation was compared with

external fixation. Measurement SDs were approximately

10 to 20 µm for overall RNFL thickness, and 5 to 9 µm

for retinal thickness. The authors found a circle diameter

of 3.4 mm to be superior; internal fixation was signifi-

cantly less variable than external fixation. Baumann et al.[62] found that the mean coefficient of varation of retinal

thickness measurements at locations outside of 500 µm

from fixation in normal eyes was 10%. The authors used

an OCT prototype characterized by a 2.5 second scan

acquisition time. Recently, Blumenthal et al. [63] evalu-

ated the CV for mean RNFL thickness in normal and

glaucomatous eyes (6.9% and 11.8% respectively) using a

Figure 3. Peripapillary retinal nerve fiber layer retardation map

and thickness plot

Peripapillary retinal nerve fiber layer (RNFL) retardation map (A) andcorresponding RNFL thickness plot (B) in the right eyes of six normal individualswith different corneal polarization axis values (18°, 27°, 37°, 52°, 59°, 76° nasallydownward from top left to bottom right). Upper and lower margins in (B)represent 95% confidence intervals. Note that peripapillary retardation andmeasured RNFL thickness increase with increasing corneal polarization axis.(Reprinted with permission: Greenfield DS, Knighton RW: Stability of cornealpolarization axis measurements for scanning laser polarimetry. Ophthalmology2001, 108:1065–1069. Figure 3).

Optic nerve and retinal nerve fiber layer analyzers in glaucoma Greenfield 73

commercially available device capable of performing

scan acquisition times in one second.

Published series of peripapillary retinal nerve fiber layer

measurement using optical coherence tomography have

sampled 100 evenly-distributed points on a 360 degree

peripapillary circular scan. Ozden et al. [64] evaluated

whether a four-fold increase in sampling density im-

proves the reproducibility of OCT measurement.

Twenty-two eyes of 22 patients (normal subjects, 3 eyes;

ocular hypertension, 2 eyes; glaucoma, 17 eyes) were

evaluated. Optical coherence tomography scanning con-

sisted of three superior and inferior quadrantic scans

(100 sampling points/ quadrant) and three circular scans

(25 points/quadrant). Retinal nerve fiber layer thickness

measurements and CV were calculated for the superior

and inferior quadrants for each sampling density tech-

nique. Normal eyes showed no difference between the

25 point/quadrant and 100 point/quadrant scans, respec-

tively. Among glaucomatous eyes, however, the CV in

25-point/quadrant scans (25.9%) was significantly higher

than that in 100-point/quadrant scans (11.9%, p = 0.01).

Sensitivity and specificity

Cross-sectional studies have compared OCT with CSLO

[65] and SLP [53] in normal, ocular hypertensive, and

glaucomatous eyes. OCT was capable of differentiating

glaucomatous from non-glaucomatous eyes, and RNFL

thickness measurements using OCT correlated with re-

tardation measurements using SLP and topographic

measurements using CSLO.

Bowd et al. [54] compared the discriminating powers of

SLP, OCT, short-wavelength automated perimetry

(SWAP), frequency-doubling technology perimetry

(FDT) for detection of early glaucoma (average visual

field mean defect –4.0 dB). The largest area under the

receiver operator characteristic (ROC) curve was found

for OCT inferior quadrant thickness, followed by the

FDT number of total deviation plot points </= 5%, SLP

linear discriminant function, and SWAP pattern SD.

Zangwill et al. [66• ] compared the ability of OCT, HRT,

and GDx to discriminate between normal eyes and eyes

with early to moderate glaucomatous visual field loss. No

significant differences were found between area under

the ROC curve and the best parameter from each instru-

ment: OCT inferior RNFL thickness, HRT mean height

contour in the inferior nasal position, and GDx linear

discriminant function).

Garcia-Sanchez et al. [29] evaluated the sensitivity and

specificity of the HRT, GDx, and OCT summary

data for detection of early to moderate glaucoma (aver-

age visual field mean defect –5.0 dB) among three

masked reviewers (see Table 1). For the OCT, sensi-

tivity and specificity ranged from 76 to 79% and 68 to

81%, respectively.

Detection of progression

Change analysis software has only recently been intro-

duced; therefore no reports have described longitudinal

change in patients with disease progression. As presently

configured, this algorithm generates a serial analysis of

RNFL thickness measurements among two OCT im-

ages, however statistical units of change probability are

not provided. Thus, true biological change cannot be

differentiated from test-retest variability.

Limitations

Currently, no statistical units of change probability are

absent from the change analysis software, therefore one

cannot differentiate biological change from measure-

ment variability by performing serial analysis of abso-

lute RNFL thickness values. Pupillary dilation is re-

quired to obtain acceptable peripapillary measurement

scans. Finally, sampling is limited to 25 A-scans per

quadrant, which may limit the ability to detect localized

change [64].

ConclusionsRecent advances in ocular imaging technology provide a

means to obtain accurate, objective, quantitative, and

reproducible structural measurements of optic disc to-

pography and RNFL thickness. Current imaging sys-

tems can differentiate between normal eyes and eyes

with mild to moderate glaucomatous optic neuropathy.

Although conflicting data exists, sensitivity and specific-

ity values approximate 70 to 80% depending upon

sample size, definition of glaucoma, and severity of glau-

comatous damage. Any one technology will have limited

usefulness as a single test to diagnose glaucoma and at

the present juncture should not be used as an indepen-

dent diagnostic screening test. However, these instru-

Figure 4. Optical coherence tomography image of a normal

eye obtained using a 3.4 mm peripapillary measurement scan

The anterior and posterior limits of the retinal nerve fiber layer (RNFL) aredemarcated using a computer algorithm (arrows) and clock hour and quadranticRNFL thickness measurements are obtained.

74 Glaucoma

ments have considerable potential for use as adjunctive

measures of glaucomatous damage along with careful

clinical and perimetric examination.

There is no uniform agreement regarding the most ap-

propriate technology for the evaluation of structural

damage in eyes with glaucomatous optic neuropathy.

Furthermore, among proponents of any given technol-

ogy, there is no consensus on the most appropriate sum-

mary measure to represent ganglion cell loss. It is im-

portant to recognize that the parameter or technology

most useful in the detection of glaucomatous damage

may vary from individual to individual and may differ

from the parameter or technology most useful for de-

tection of glaucomatous change. The most appropriate

measure(s) of disease detection will unlikely be the

most sensitive indicator of glaucomatous change. At

the present time, limited information exists regarding

the relation between glaucomatous progression and

RNFL/topographic measures.

Currently available imaging technologies hold consider-

able promise for detection of glaucomatous change.

Methods for change detection exist but have not been

prospectively validated in large populations. Moreover,

new strategies for detection of progressive structural

change need to be validated against accepted measures

of structural (stereoscopic disc photography) and func-

tional (psychophysical) change. Statistical units of

change probability are essential to differentiate true bio-

logical change from variability (eg, microsaccades during

fixation, vessel pulsations, instrument or operator-

induced variability). A significant challenge to the inves-

tigator has been the reality that technology improves

with time. Rapidly evolving hardware and software re-

sults in alteration of baseline measurements. This has

produced instability in longitudinal data collection and

has limited, in part, our ability to critically evaluate the

efficacy of these instruments to detect structural change

over time. Presently, it is unclear whether automated

detection of structural change meets or exceeds current

standard of care measures.

In summary, each ocular imaging technology has specific

advantages and disadvantages. One instrument may not

be best for all purposes and all patients, and different

analysis strategies may not agree. Because measure-

ment reproducibility is high, each technology holds

promise for improving our ability to detect glaucoma-

tous change. As with perimetry, it is not recommended

that isolated clinical decisions be based solely upon

ocular imaging results. Clinical correlation should be

performed and treatment recommendations should

be individualized.

Acknowledgments

Supported in part by the New York Community Trust, New York, New York; TheKessel Foundation, Bergenfield, New Jersey; The Boyer Foundation, Melbourne,FL; and NIH Grant R01-EY08684, Bethesda, Maryland. The author has no propri-etary interest in any of the products or techniques described in this manuscript.

References and recommended reading

Papers of particular interest, published within the annual period of review,have been highlighted as:• Of special interest•• Of outstanding interest

1 Sommer A, Miller NR, Pollack I, et al.: The nerve fiber layer in the diagnosis ofglaucoma. Arch Ophthalmol 1977, 95:2149.

2 Quigley HA, Dunkelberger GR, Green WR: Retinal ganglion cell atrophy cor-related with automated perimetry in human eyes with glaucoma. Am J Oph-thalmol 1989, 107:453–464.

3 Quigley HA: Better methods in glaucoma diagnosis. Arch Ophthalmol 1985,103:186.

4 Sommer A, Katz J, Quigley HA, et al.: Clinically detectable nerve fiber layeratrophy precedes the onset of glaucomatous field loss. Arch Ophthalmol1991, 109:77–83.

5 Lusky M, Bosem ME, Weinreb RN: Reproducibility of optic nerve head to-pography measurements in eyes with undilated pupils. J Glaucoma 1993,2:104–109.

6 Weinreb RN, Dreher AW, Bille JF: Quantitative assessment of the optic nervehead with the laser tomographic scanner. Int Ophthalmol 1989, 13:25.

7 Brigatti L, Weitzman M, Caprioli J: Regional test-retest variability of confocalscanning laser tomography. Am J Ophthalmol 1995, 120:433–440.

8 Zangwill L, Schakiba S, Caprioli J, et al.: Agreement between clinicians and aconfocal scanning laser ophthalmoscope in estimating cup-to-disc ratios. AmJ Ophthalmol 1995; 119:415–421.

9 Dreher AW, Reiter K, Weinreb RN: Spatially resolved birefringence of theretinal nerve fiber layer assessed with a retinal ellipsometer. Applied Optics1992, 31:3730–3749.

10 Weinreb RN, Shakiba S, Zangwill L: Scanning laser polarimetry to measurethe nerve fiber layer of normal and glaucomatous eyes. Am J Ophthalmol1995, 119:627–636.

11 Huang D, Swanson EA, Lin CP, et al.: Optical coherence tomography. Sci-ence 1991 254:1178–1181.

12 Izatt JA, Hee MR, Swanson EA, et al.: Micrometer-scale resolution imaging ofthe anterior eye in vivo with optical coherence tomography. Arch Ophthalmol1994, 112:1584–1589.

13 Hee MR, Izatt JA, Swanson EA, et al.: Optical coherence tomography of thehuman retina. Arch Ophthalmol 1995, 113:325–332.

14 Weinreb RN, Lusky M, Bartsch D, et al.: Effect of repetitive imaging on topo-graphic measurements of the optic nerve head. Arch Ophthalmol 1993,111:636–638.

15 Mikelberg FS, Wijsman K, Schulzer M: Reproducibility of topographic param-eters obtained with the Heidelberg Retina Tomograph. J Glaucoma 1993,2:101–103.

16 Dreher AW, Tso PC, Weinreb RN: Reproducibility of topographic measure-ments of the normal and glaucomatous optic nerve head with the laser tomo-graphic scanner. Am J Ophthalmol 1991, 111:221.

17 Zangwill L, Irak I, Berry CC, et al.: Effect of cataract and pupil size on imagequality with confocal scanning laser ophthalmoscopy. Arch Ophthalmol1997, 115:983–990.

18 Mikelberg FS, Parfitt CM, Swindale NV, et al.: Ability of the Heidelberg retinatomograph to detect early glaucomatous field loss. J Glaucoma 1995,4:242–247.

19 Bartz-Schmidt KU, Sengersdorf A, Esser P, et al.: The cumulative normalisedrim/disc area ratio curve. Graefes Arch Clin Exp Ophthalmol 1006, 234:227–231.

20 Asawaphureekorn S, Zangwill L, Weinreb RN: Ranked-segment distributioncurve for interpretation of optic nerve topography. J Glaucoma 1996,5:79–90.

21 Wollstein G, Garway-Heath DF, Hitchings RA: Identification of early glau-coma cases with the scanning laser ophthalmoscope. Ophthalmology 1998,105:1557–1563.

•22 Wollstein G, Garway-Heath DF, Fontana L, et al.: Identifying early glaucoma-

tous changes: comparison between expert clinical assessment of optic discphotographs and confocal scanning ophthalmoscopy. Ophthalmology 2000,107:2272–2277.

This study found the Heidelberg Retinal Tomograph to be superior to clinical evalu-ation of optic disc stereophotographs but was limited by the use of disparate ob-servers with limited levels of agreement.

23 Iester M, Mikelberg FS, Drance SM: The effect of optic disc size on diagnosticprecision with the Heidelberg retina tomograph. Ophthalmology 1997,104:545–548.

Optic nerve and retinal nerve fiber layer analyzers in glaucoma Greenfield 75

24 Bathija R, Zangwill L, Berry CB, et al.: Detection of early glaucomatous struc-tural damage with confocal scanning laser tomography. J Glaucoma 1998,7:121–127.

25 Caprioli J, Park HJ, Ugurlu S, et al.: Slope of the peripapillary nerve fiber layersurface in glaucoma. Invest Ophthalmol Vis Sci 1998, 39:2321–2328.

26 Gundersen KG, Asman P: Comparison of ranked segment analysis (RSA)and cup to disc ratio in computer-assisted optic disc evaluation. Acta Oph-thalmol Scand 2000, 78:137–141.

27 Swindale NV, Stjepanovic G, Chin A, et al.: Automated analysis of normal andglaucomatous optic nerve head topography images. Invest Ophthalmol VisSci 2000, 41:1730–1742.

•28 Miglior S, Casula M, Guareschi M, et al.: Clinical ability of Heidelberg Retinal

Tomograph examination to detect glaucomatous visual field changes. Oph-thalmology 2001, 2001:1621–1627.

In a broad clinical setting of 359 normal, ocular hypertensive, or glaucomatouseyes, this study found that the Heidelberg Retinal Tomograph lacked sensitivity andspecificity when using commercially available ranked-segment distribution curves,and Mikelberg’s multivariate discriminant analysis, respectively.

•29 Sanchez-Galeana C, Bowd C, Blumenthal EZ, et al.: Using optical imaging

summary data to detect glaucoma. Ophthalmology 2001, 108:1812–1818.This report evaluated the sensitivity and specificity of ocular imaging summary datausing three experienced observers with high levels of agreement and found similardiscriminating powers among HRT, GDx, and OCT.

30 Anderson DR, Chauhan B, Johnson C, et al.: Criteria for progression of glau-coma in clinical management and in outcome studies. Am J Ophthalmol 2000,130:827–829.

••31 Chauhan BC, Blanchard JW, Hamilton DC, et al.: Technique for detecting

serial topographic changes in the optic disc and peripapillary retina usingscanning laser tomography. Invest Ophthalmol Vis Sci 2000, 41:775–782.

This study describes a novel change analysis algorithm for differentiating test-retest variability from statistically significant change in optic disc topography and iscurrently employed within the latest version of HRT software.

32 Kamal DS, Viswanathan AC, Garway-Heath DF, et al.: Detection of optic discchange with the Heidelberg Retina Tomograph before confirmed visual fieldchange in ocular hypertensives converting to early glaucoma. Br J Ophthalmol1999, 83:290–294.

33 Lusky M, Morsman D, Weinreb RN: Effects of intraocular pressure on opticnerve head topography. Curr Opin Ophthalmol 1993, 4:40–44.

34 Chauhan BC, MacDonald CA: Influence of time separation on variability es-timates of topographic measurements with confocal scanning laser tomog-raphy. J Glaucoma 1995, 4:189–193.

35 Lesk MR, Spaeth GR, Azuara-Blanco A, et al.: Reversal of optic disc cuppingafter glaucoma surgery analyzed with a scanning laser tomograph. Ophthal-mology 1999, 106:1013–1018.

36 Irak I, Zangwill L, Garden V, et al.: Change in optic disk topography aftertrabeculectomy. Am J Ophthalmol 1996, 122:690–695.

37 Lee BL, Zangwill L, Weinreb RN: Change in optic disc topography associatedwith diurnal variation of intraocular pressure. J Glaucoma 1999; 8:221–223.

38 Tjon-Fo-Sang MJ, Lemij HG: The sensitivity and specificity of nerve fiber layermeasurements in glaucoma as determined with scanning laser polarimetry.Am J Ophthalmol 1997, 123:62–69.

39 Niessen AGJE, Van Den Berg TJTP, Langerhorst CT, et al.: Retinal nerve fiberlayer assessment by scanning laser polarimetry and standardized photogra-phy. Am J Ophthalmol 1996, 121:484–493.

40 Hoh ST, Ishikawa H, Greenfield DS, et al.: Peripapillary nerve fiber layer thick-ness measurement reproducibility using scanning laser polarimetry. J Glau-coma 1998, 7:12–15.

41 Reiter K, Dreher A, inventors; Laser Diagnostic Technologies, Inc., assignee.Eye examination apparatus employing polarized light probe. 1998 Aug 4.United States Patent No. 5,787,890.

42 Dreher A, Reiter K, inventors; Laser Diagnostic Technologies, Inc., assignee.Retinal eye disease diagnostic system. 1994 April 19. United States PatentNo. 5,303,709.

43 Knighton RW, Huang W-R, Greenfield DS: Linear birefringence measured inthe central cornea of a normal population. Invest Ophthalmol Vis Sci 2002,43:82–86.

44 Greenfield DS, Knighton RW, Feuer W, et al.: Correction for corneal polar-ization axis improves the discriminating power of scanning laser polarimetry.Invest Ophthalmol Vis Sci 2001; 42:S314.

45 Hoh ST, Greenfield DS, Liebmann JM, et al.: Effect of pupillary dilation onretinal nerve fiber layer thickness measurement using scanning laser polarim-etry. J Glaucoma 1999, 8:159–163.

46 Chi QM, Tomita G, Inazumi K, et al.: Evaluation of the effect of aging on theretinal nerve fiber layer thickness using scanning laser polarimetry. J Glau-coma 1995, 4:406–413.

47 Swanson WH, Lynn JR, Fellman RL, et al.: Inter-operator variability in imagesobtained by laser polarimetry of the nerve fiber layer. J Glaucoma 1995,4:414–418.

48 Zangwill L, Berry CA, Garden VS, et al.: Reproducibility of retardation mea-surements with the Nerve Fiber Analyzer II. J Glaucoma 1997, 6:384–389.

49 Tjon-Fo-Sang MJH, van Strik R, de Vries J, et al.: Improved reproducibility ofmeasurements with the Nerve Fiber Analyzer. J Glaucoma 1997, 6:203–211.

50 Waldock A, Potts MJ, Sparrow JM, et al.: Clinical evaluation of scanning laserpolarimetry: Intraoperator reproducibility and design of a blood vessel re-moval algorithm. Br J Ophthalmol 1998, 82:252–259.

51 Weinreb RN, Zangwill L, Berry CC, et al.: Detection of glaucoma with scan-ning laser polarimetry. Arch Ophthalmol 1998, 116:1583–1589.

52 Garcia-Sanchez J, Garcia-Feijoo J, Arias-Puente A, et al.: Accuracy of theGDx system for the diagnosis of glaucoma. Invest Ophthalmol Vis Sci 1998,38:933.

53 Hoh ST, Greenfield DS, Mistlberger A, et al.: Optical coherence tomographyand scanning laser polarimetry in normal, ocular hypertensive, and glaucoma-tous eyes. Am J Ophthalmol 2000, 129:129–135.

54 Bowd CA, Zangwill LM, Berry CC, et al.: Detecting early glaucoma by as-sessment of retinal nerve fiber layer thickness and visual function. Invest Oph-thalmol Vis Sci 2001, 42:1993–2003.

55 Colen TP, Van Everdingen JAM, Lemij HG: Axonal loss in a patient with an-terior ischemic optic neuropathy as measured with scanning laser polarim-etry. Am J Ophthalmol 2000, 130:847–850.

56 Medeiros FA, Susanna R: Retinal nerve fiber layer loss after traumatic opticneuropathy detected by scanning laser polarimetry. Arch Ophthalmol 2001,119:920–921.

••57 Greenfield DS, Huang X-R, Knighton RW: Effect of corneal polarization axis

on assessment of retinal nerve fiber layer thickness by scanning laser polar-imetry. Am J Ophthalmol 2000, 129:715–722.

This report describes the effect of corneal birefringence upon RNFL thicknessdeterminations using the GDx nerve fiber analyzer and outlines the optical limita-tions of using a fixed corneal compensator to neutralize anterior segment polariza-tion. A novel method for estimating corneal birefringence is described using macu-lar birefringence characteristics.

58 Greenfield DS, Knighton RW: Stability of corneal polarization axis measurementsfor scanning laser polarimetry. Ophthalmology 2001, 108:1065–1069.

59 Hoh ST, Greenfield DS, Ishikawa H, et al.: Factors affecting image acquisitionduring scanning laser polarimetry. Ophthalmic Surg and Lasers 1998,29:545–551.

60 Fercher AF, Hitzenberger CK, Drexler W, et al.: In vivo optical coherencetomography. Am J Ophthalmol 1993, 116:113–114.

61 Schuman JS, Pedut-Kloizman T, Hertzmark E, et al.: Reproducibility of nervefiber layer thickness measurements using optical coherence tomography.Ophthalmology 1996, 103:1889–1898.

62 Baumann M, Gentile RC, Liebmann JM, et al.: Reproducibility of retinal thick-ness measurements in normal eyes using optical coherence tomography.Ophthalmic Surg Lasers 1998, 29:280–285.

63 Blumenthal EZ, Williams JM, Weinreb RN, et al.: Reproducibility of nerve fiberlayer thickness measurements by use of optical coherence tomography. Oph-thalmology 2000, 107:2278–2282.

64 Ozden RG, Ishikawa HI, Liebmann JM, et al.: Increasing sampling densityimproves optical coherence tomography measurement reproducibility. JGlaucoma 1999, 8:238–241.

65 Mistlberger A, Liebmann JM, Greenfield DS, et al.: Heidelberg retina tomog-raphy and optical coherence tomography in normal, ocular hypertensive andglaucomatous eyes. Ophthalmology 1999, 106:2027–2032.

•66 Zangwill LM, Bowd C, Berry CC, et al.: Discriminating between normal and

glaucomatous eyes using the Heidelberg Retina Tomograph, GDx Nerve Fi-ber Analyzer, and Optical Coherence Tomograph. Arch Ophthalmol 2001,119:985–993.

This report provides a comparison of the ability of OCT, GDx, and HRT to discrimi-nate between healthy eyes and eyes with mild to moderate glaucoma and found nodifferences between the best parameter for each instrument using receiver opera-tor characteristic curves.

67 Kruse FE, Burk ROW, Volcker H-E, et al.: Reproducibility of topographic mea-surements of the optic nerve head with laser tomographic scanning. Ophthal-mology 1989, 96:1320–1324.

68 Choplin NT, Lundy DC, Dreher AW: Differentiating patients with glaucomafrom glaucoma suspects and normal subjects by nerve fiber layer assessmentwith scanning laser polarimetry. Ophthalmology 1998, 105:2068–2076.

76 Glaucoma