Download Reproducibility of photogrammetric optic disc cup

Reproducibility of Photogramnnetric Optic
Disc Cup Measurements
Takenori Takamoro and Bernard Schwartz
To detect early glaucoma-associated changes in the optic disc cup, the authors analyzed reproducibility
of optic disc cup measurements by photogrammetry of three age-matched groups: 10 normal, 10
ocular hypertensive, and 10 glaucomatous subjects. For each eye, three simultaneous stereophotographs
were taken on the same day and one measurement was made for each stereophotograph. In order to
reduce the effects of refractive error of the eye and photographic magnification, the authors computed
cup parameters as relative measurements, ie, a ratio to the optic disc area. The median coefficients
of variation for glaucomatous eyes were 4.5% for cup volume/disc area, 6.1% for cup depth/disc area,
and 4.7% for cup area/disc area. Coefficients of variation of normal and ocular hypertensive subjects
were slightly larger than those of glaucomatous subjects. Coefficients of variation generally were
larger for quadrants than for total disc area. These findings suggest that the photogrammetric
technique is highly reproducible and sensitive and could be a valuable quantitative technique for the
clinical study of glaucoma. Invest Ophthalmol Vis Sci 26:814-817, 1985
A three-dimensional depression below the retinal
surface of the optic nerve is referred to as cupping of
the optic disc. It appears to correlate with the development of glaucomatous visual field defects,1 and an
increase in its volume may precede a glaucomatous
visual field defect.2 Reproducible measurement of
optic disc cupping may be effective to detect an early
stage of glaucoma. As an objective and quantitative
method to evaluate the cupping of the optic disc,
photogrammetric techniques using simultaneous stereophotographs have been applied for measurement
of topographic cup parameters3"9 and for detection
of relative changes in the dimensions of the optic
disc cup. 10
In this investigation, we determined the reproducibility of our methods of photogrammetric measurements of the optic disc cup parameters by measuring
normal, ocular hypertensive, and open-angle glaucomatous subjects.
England Medical Center as follows: 10 normal subjects
[intraocular pressure (IOP) of <21 mmHg in both
eyes on two or more independent examinations,
normal visual fields]; 10 untreated ocular hypertensive
subjects (IOP of >21 mmHg in at least one eye on
at least two independent examinations, normal visual
fields); and 10 subjects with open-angle glaucoma
(IOP of >21 mmHg in at least one eye on at least
two independent examinations, with visual field and
optic disc changes characteristic of glaucoma). Informed human consent was obtained prior to undertaking the study.
Both eyes of all subjects had simultaneous stereophotographs taken three times on each visit. We used
the Donaldson stereoscopic fundus camera" with 3X
photographic magnification, 4-mm camera aperture
size, and Kodak Ektachrome 200 color film, all of
which contributed to optimum reproducibility.7 In
order to minimize image distortions, which were
larger at the periphery than at the center of the film
frame,12 the image of the optic disc was centered on
the film frame. We adjusted for camera alignment,
focusing, and centering of optic disc image at each
exposure. We selected one eye from each subject (five
left eyes and five right eyes from each group). With
the stereoplotter Kern PG2 (Kern Co., Ltd.; Aarau,
Switzerland) and a pair of stereophotographs, we
digitized each optic disc in three dimensions, including
margins, with well-distributed points (about 300).
Each of three stereopairs of each eye was digitized
separately. The digitized data were then stored in the
disc memory of a Digital Equipment Corporation
(Maynard, MA) VAX 11/780 computer and were
Materials and Methods
Thirty age-matched subjects were randomly selected
from the ophthalmology clinic of the Tufts-New
From the Department of Ophthalmology, New England Medical
Center and Tufts University School of Medicine, Boston, Massachusetts.
Supported in part by grants EY-00936 and EY-03661 from the
National Eye Institute, National Institutes of Health, Bethesda,
Maryland.
Submitted for publication: April 5, 1984.
Reprint requests: Takenori Takamoto, PhD, Department of
Ophthalmology, New England Medical Center, 171 Harrison Avenue, Boston, MA 02111.
814
Downloaded From: http://iovs.arvojournals.org/pdfaccess.ashx?url=/data/journals/iovs/933352/ on 06/14/2017
No. 6
815
REPRODUCIDILITY OF OPTIC DISC PHOTOGRAAAMETRY / ToKomoro ond Schworrz
Table 1. Reproducibility of photogrammetric measurements of the optic disc cup"
Ocular hypertensives (n = 10)
Normals (n = 10)
Cup volume/disc
area
Superior
Inferior
Nasal
Temporal
Total
Cup depth/disc
area
Superior
Inferior
Nasal
Temporal
Total
Cup area/disc
area
Superior
Inferior
Nasal
Temporal
Total
Glaucomas (n = 10)
Median
SD
CV (%)
Median
SD
CV (%)
Median
SD
CV (%)
0.151
0.151
0.080
0.189
0.151
0.012
0.012
0.015
0.024
0.010
7.2
7.2
12.5
11.9
7.7
0.214
0.178
0.138
0.194
0.173
0.019
0.014
0.025
0.026
0.018
6.5
8.8
17.5
12.1
7.9
0.386
0.389
0.326
0.329
0.352
0.037
0.016
0.026
0.033
0.018
11.9
5.7
9.3
0.178
0.146
0.162
0.154
0.165
0.010
0.014
0.021
0.017
0.009
6.5
7.2
10.9
9.8
4.6
0.182
0.188
0.189
0.187
0.200
0.018
0.016
0.019
0.016
0.012
11.9
9.5
9.7
9.5
6.7
0.240
0.206
0.248
0.229
0.238
0.016
0.019
0.026
0.019
0.015
8.7
10.0
13.8
10.1
6.1
0.491
0.479
0.328
0.600
0.470
0.029
0.042
0.042
0.074
0.026
5.9
9.1
12.8
12.0
5.2
0.620
0.429
0.379
0.568
0.486
0.037
0.048
0.054
0.050
0.037
6.0
9.8
19.2
8.8
7.6
0.713
0.758
0.693
0.690
0.717
0.031
0.032
0.027
0.041
0.027
4.9
8.9
4.5
3.9
4.4
5.0
4.7
* The medians, standard deviations (SD), and the coefficients of variation
(CV) of the cup parameters were medians (50th percentile) for the three groups:
10 normal, 10 ocular hypertensive, and 10 open-angle glaucomatous subjects.
used to compute cup parameters across the entire
disc and for each disc quadrant by the radial section
method.6 This method is used to: (1) define the optic
disc plane as a reference plane of horizontal-vertical
coordinates; (2) define the depth axis at the gravity
center of the disc area; and (3) compute the size of
an averaged radial section on each quadrant.
Cup edge is a change of slope from the cup wall
to the neural rim. The cup edge was digitized as a
series of discrete points. The disc edge was similarly
digitized. The cup area was computed as an area
delineated by cup edge, and the disc area was computed as an area delineated by disc edge. Cup depth
was computed as the distance from the bottom of
the cup to the plane defined by the disc edge. The
cup volume was computed as the depressed area of
the optic disc surface relative to the plane defined by
the disc edge. Photographic magnification and refractive error of the eye affect the measurement of the
area and depth (and therefore the volume) of the
optic disc cup. 13 In order to reduce these effects, we
computed cup parameters as relative to the optic disc
area; that is, cup volume/disc area, cup depth/disc
area, and cup area/disc area.
The mean of each cup parameter and its standard
deviation and coefficient of variation, that is, the
standard deviation divided by the mean and expressed
as a percentage, were computed from three measurements (one measurement of each of three photographs
of the same eye) for quadrants and the total disc.
Medians of these means, standard deviations, and
coefficients of variation were computed for the group
of normal, ocular hypertensive, and open-angle glaucomatous subjects.
Results
The median for the cup parameters, the standard
deviations, and coefficients of variation for each
group are listed in Table 1. Generally the coefficients
of variation were slightly larger in normal and ocular
hypertensive subjects than in glaucomatous subjects
except for cup depth/disc area.
Comparing the coefficients of variation among
quadrants for each cup parameter, the largest was
usually found in nasal quadrants.
In the total disc area, the cup volume/disc area in
glaucomatous eyes was more than twice as large as
in the normal and ocular hypertensive eyes. The
median cup depth/disc area in open-angle glaucoma
was about 20% larger than in ocular hypertensive
eyes and more than 40% larger than in normal eyes.
The median cup area/disc area of glaucomatous eyes
was about 50% larger than in normal and ocular
hypertensive eyes.
In the nasal quadrant, the median cup volume/
disc area of glaucoma was more than four times as
large as in the normal eyes and more than twice as
large as in the ocular hypertensive eyes. The median
value for the ocular hypertensive eyes was about 70%
larger than for the normal eyes.
Spearman correlations between mean and coeffi-
Downloaded From: http://iovs.arvojournals.org/pdfaccess.ashx?url=/data/journals/iovs/933352/ on 06/14/2017
816
INVESTIGATIVE OPHTHALMOLOGY 6 VISUAL SCIENCE / June 1985
Vol. 26
Table 2. Comparison of reported reproducibility of optic cup measurements
Coefficient of variation
Author
Method
14
No. of eyes
Volume
Depth
Area
3.1 (median)
6.1 (median)
Holm and Krakau
(1970)
Optical sectioning
(photograph with
projected section
lines)
1N
1G
7.3, 5.6
8.8
Krakau and Torlegard15
(1972)
Optical sectioning
(photograph with
projected section
lines)
1N
2.2
Rosenthal et al3 (1977)
Digital photogrammetry
(simultaneous
stereophotographs)
3*
7.8 (median)
Takamoto et al6 (1979)
Stereophotogrammetry
(simultaneous
stereophotographs)
3N
Takamoto and Schwartz7
(1979)
Stereophotogrammetry
(simultaneous
stereophotographs)
5N
9.6 (average)
4.2 (average)
8.4 (average
cup/disc)
Krohn et al5 (1979)
Stereophotogram metry
(simultaneous
stereophotographs)
3 N, 5 G
8.0 (median)
7.0 (median)
8.0 (median)
Takamoto and Schwartz
(1984)
Stereophotogrammetry
(simultaneous
stereophotographs)
10 N
10 OH
10G
7.7f
7.9f
4.5f
4.6*
6.7*
6.1*
5.2§
7.6§
4.7§
N: normal eye: G: glaucomatous eye; OH: ocular hypertensive eye.
* Diagnoses were not available for these eyes.
t Median cup volume/disc area.
cient of variation of cup parameters of total disc area
were calculated. No significant correlations at a P
level of <0.05 were found for cup depth/disc area in
any of the three groups (normal, ocular hypertensive,
or open-angle glaucomatous subjects) or in the total
population. Borderline negative correlations (the larger
the mean, the smaller the coefficient of variation)
were found for cup area/disc area (r = -0.3157, P
= 0.089) for the total population (n = 30) and for
cup volume/disc area (r = -0.6000, P = 0.074) for
normal subjects (n = 10).
Discussion
The object of our study was to estimate the reproducibility of the cup parameters of the optic disc by
photogrammetry with the use of simultaneous stereophotographs taken under routine clinical conditions.
Therefore, none of 90 pairs of stereophotographs (30
eyes X three stereophotographs) were rejected because
of poor image quality or poor centration of the image
of the optic disc on the film format.
The reproducibility of cup parameter measurements
as determined by the coefficients of variation for the
quadrants usually were greater than those for the
whole disc; medians of the coefficient of variation for
the total disc for all cup parameters ranged from
4.5% to 7.9%, while medians of the coefficient of
3.0 (average)
^ Median cup depth/disc area.
§ Median cup area/disc area.
variation for the quadrants ranged from 3.9% to
19.2%. Two major factors may have contributed to
this. First, the number of measured points used to
compute the quadrant area was about one-fourth of
that of total disc area. This cannot be changed since
we use the same amount of digitized data for cup
parameters of quadrants and total disc area. Second,
the consistency of the process of quadrant subdivision
depends on the vertical (superior-inferior) axis of the
optic disc relative to that of the camera. This error
can be reduced by careful positioning of the subject's
head but appears to be difficult to avoid completely.
No significant correlation was found between mean
and coefficient of variation of cup parameters of total
disc area or of any quadrant; that is, reproducibility
was the same for large and small optic discs.
Reproducibility of cup parameters was significantly
better3 or similar5 with simultaneous stereophotographs than with sequential stereophotographs, which
may be affected by movements of the camera or of
the patient's eye between exposures. Therefore, we
did not include sequential stereophotographs in this
investigation of reproducibility.
A comparison of the results of the present study
and those of several other investigations of reproducibility of optic disc cup measurements is presented in
Table 2. Holm and Krakau14 and Krakau and Tor-
Downloaded From: http://iovs.arvojournals.org/pdfaccess.ashx?url=/data/journals/iovs/933352/ on 06/14/2017
No. 6
REPRODUCIBILITY OF OPTIC DISC PHOTOGRAMMETRY / Tokomoro ond Schworrz
legard15 projected slit beams of light onto optic discs
and used a fundus camera to photograph the discs.
Results of optical sectioning in studies based on one
or two eyes had coefficients of variation that ranged
from 2.2% to 8.8%. Using digital photogrammetry,
which is a computer image analysis of stereophotographs of the optic disc, Rosenthal and colleagues3
achieved results similar to ours, but from a smaller
sample size.
In our study with Marzan6 (Table 2), cup depth
was computed as the distance between two points, a
point on the cup rim and a point on the cup bottom.
In the present study, cup depth was computed as the
distance between two planes, with one plane representing the top of the retinal surface and the other
representing the bottom of the cup. The point-topoint method yielded better reproducibility of cup
depth measurement than the current investigation
but selected a smaller sample of points for depth
measurement. In another study,7 we compensated for
differences in magnification between photographs obtained at different times by using several anatomic
reference points on the disc; this study demonstrated
reproducibility generally poorer than that found in
the present study. Krohn, Keltner, and Johnson, 5
using a Donaldson stereofundus camera and stereophotogram metric methods to measure normal and
glaucomatous eyes that had been masked for diagnosis,
also achieved poorer reproducibility than our present
study (Table 2).
The advantages of our methods and results are
apparent in Table 2. Compared to other investigators,
we used a larger series in three groups and achieved
smaller coefficients of variation.
Using model eyes, Rosenthal, Falconer, and Pieper16
reported on the sensitivity of photogrammetric measurements with digital image processing. They compared actual cup depths of a model eye with photogrammetric measurements, and the differences were
less than 260 iim for the cup depth of up to 3.18
mm. In our similar study,12 calibrated cup depths of
a model eye were compared with measurements
obtained by photogram metric methods; the differences
were less than 200 jum when the image of the fundus
of the model eye was located in the central part of
the film frame.
Our results suggest that the photogram metric technique is highly reproducible and sensitive in detecting
cup changes and could be beneficial in a clinical
setting. The clinician applying this technique could
send stereophotographs to a photogrammetry center
for optic disc cup measurement, which presently
takes only about 10 min for one optic disc and is
followed by computations and plottings of cup parameters by computer programs. The further development of computerized image analysis for photo-
817
grammetric measurements will allow for even more
efficiency and cost-effectiveness.
Key words: optic disc cup volume, glaucoma, ocular hypertension, stereophotograph, photogrammetry
Acknowledgments
The authors thank Judith Barton, MS, for statistical
analysis; Susan Glick, MS, for editing the manuscript;
Eileen Walsh, for secretarial help; and Benedicto Arevalo,
for assistance in preparation of the manuscript.
References
1. Sommer A, Pollack I, and Maumenee AE: Optic disc parameters
and onset of glaucomatous field loss. I. Methods and progressive
changes in disc morphology. Arch Ophthalmol 97:1444, 1979.
2. Pederson JE and Anderson DR: The mode of progressive disc
cupping in ocular hypertension and glaucoma. Arch Ophthalmol
98:490, 1980.
3. Rosenthal AR, Kottler MS, Donaldson DD, and Falconer DG:
Comparative reproducibility of the digital photogrammetric
procedure"utilizing three methods of stereophotography. Invest
Ophthalmol Vis Sci 16:54, 1977.
4. Johnson CA, Keltner JL, Krohn MA, and Portney GL: Photogrammetry of the optic disc in glaucoma and ocular hypertension with simultaneous stereo photography. Invest Ophthalmol Vis Sci 18:1252, 1979.
5. Krohn MA, Keltner JL, and Johnson CA: Comparison of
photographic techniques and films used in stereophotogrammetry of the optic disc. Am J Ophthalmol 88:859, 1979.
6. Takamoto T, Schwartz B, and Marzan GT: Stereo measurements
of the optic disc. Photogrammetric Engineering and Remote
Sensing 45:79, 1979.
7. Takamoto T and Schwartz B: Topographic parameters of the
optic disc by the radial section method. In Proceedings of the
American Society of Photogrammetry, Falls Church, VA, The
Society, 1979, pp. 238-251.
8. Betz P, Camps F, Collignon-Brach J, and Weekers R: Photographie stereoscopique et photogrammetrie de 1'excavation
physiologique de la papille. J Fr Ophtalmol 4:193, 1981.
9. Betz P, Camps F, Collignon-Brach J, Lavergne G, and Weekers
R: Biometric study of the disc cup in open-angle glaucoma.
Graefes Arch Clin Exp Ophthalmol 218:70, 1982.
10. Takamoto T and Schwartz B: Photogrammetric measurements
of the optic disc in glaucoma. International Archives of Photogrammetry 23(B5):732, 1980.
11. Donaldson D, Prescott R, and Kennedy S: Simultaneous
stereoscopic fundus camera utilizing a single optical axis. Invest
Ophthalmol Vis Sci 19:289, 1980.
12. Takamoto T, Kennedy S, and Schwartz B: Image distortions
of Donaldson stereoscopic fundus camera. In Proceedings of
The American Society of Photogrammetry, Falls Church, VA,
The Society, 1981, pp. 325-334.
13. Takamoto T and Schwartz B: Stereo measurements of the
optic disc cup shape: volume profile method. In Proceedings
of the American Society of Photogrammetry, Falls Church,
VA, The Society, 1984, pp. 352-358.
14. Holm O and Krakau CET: A photographic method for measuring the volume of papillary excavation. Ann Ophthalmol 1:
327, 1970.
15. Krakau CET and Torlegard K: Comparison between stereo
and slit like image photogrammetric measurements of the optic
disc. Acta Ophthalmol 50:863, 1972.
16. Rosenthal AR, Falconer DG, and Pieper I: Photogrammetry
experiments with a model eye. Br J Ophthalmol 64:881, 1980.
Downloaded From: http://iovs.arvojournals.org/pdfaccess.ashx?url=/data/journals/iovs/933352/ on 06/14/2017