Abstract
purpose. To develop and validate a digital imaging and analysis technique for
assessing the extent of posterior capsular opacification after cataract
surgery.
methods. Retroillumination images of the posterior capsule were obtained by
using a digital camera mounted on a slit lamp. The images were analyzed
using an available image analysis software program. The image
acquisition and analysis techniques were tested for face validity,
reproducibility, and the ability to detect progression of capsular
opacity over time.
results. Digital retroillumination images were obtained without patient
discomfort. Automated analysis of images correlated well with clinical
grading both at slit lamp examination and when looking at the images
themselves (Spearman’s correlation coefficient >0.7). Analysis of
images taken at different times showed high reproducibility (intraclass
correlation >0.9), and the system was able to identify progression of
capsular opacity over a 2-year period with a mean increase of 15.8% in
progressors versus an increase of 0.6% in nonprogressors
(P < 0.05).
conclusions. Digital retroillumination images of the posterior capsule can be
obtained reliably, and automated analyses correlate well with clinical
assessment. The system presented here uses commercially available
instruments and software, and it is practical for use in longitudinal
studies of posterior capsule opacification. It is reliable, easy to
use, and can detect small changes in the percentage area covered by
posterior capsule opacification over time.
More than 1 million cataract surgeries are performed annually in
the United States, and virtually all are performed using extracapsular
extraction techniques in which the posterior capsule of the lens is
left intact. Lens epithelial cells that remain adherent to the lens
capsule have the potential to proliferate and transform into
fibroepithelial sheets, which may lead to significant posterior
capsular opacification (PCO) and necessitate laser capsulotomy to
restore vision.
1 This procedure, Nd:YAG capsulotomy, is
associated with ocular morbidity and considerable costs. It has been
estimated that approximately 25% of cataract surgical patients in the
United States undergo this treatment within 2 years of cataract
surgery.
2 3 Nd:YAG capsulotomy is now the second most
commonly billed procedure among Medicare beneficiaries. In 1995,
approximately 650,000 Nd:YAG capsulotomies were performed on Medicare
beneficiaries at charges for that year of approximately $300 million
(Earl Steinberg, Medicare Cataract Surgery records, 1995; personal
communication, April 18, 1998).
Although the Nd:YAG capsulotomy rate is important for estimating health
care costs, the reported rate is a poor surrogate for studying the
biology of PCO. The rate of capsulotomy varies by
surgeon,
4 geographic region,
5 and patient
needs. A reliable and valid method of measuring the degree of PCO is
needed to assess risk factors and potential mechanisms for posterior
capsule opacification. Desirable features of a grading scheme for PCO
would include high reproducibility and validity, sensitivity to PCO
progression, ease of use, straightforward archiving and recall of
images, suitability for standard image analysis strategies, instant
feedback on image quality to allow for correlation between the image
and the clinical picture, and generalizability to other investigators
for facilitation of future research.
To date, there has been little published on the grading of
PCO.
6 7 8 Lasa et al.
6 used a Scheimpflug slit
lamp camera (Anterior Eye Segment Analysis System, EAS-1000; Nidet,
Gamagori, Japan) to capture a still image of the posterior
capsule and image analysis software to calculate density and thickness.
Although this technique could, on average, separate patients clinically
thought to have hazy capsules, it is limited by the small area of the
slit beam that is assessed, the high cost of the instruments used, and
the inability of the system to assess total area of the capsule covered
by opacity without using multiple images. In addition, its reliability
and ability to detect progression over time have not been reported.
Hayashi et al.
7 used the Scheimpflug to assess the central
3 mm of the posterior capsule to determine “density,” measured in
computer-compatible tapes. They found their measurements to be
correlated with visual acuity but have not yet documented the
reproducibility of their approach or the ability to detect progression
over time. In addition, their technique only images four slices through
the capsule, raising the possibility that capsular opacities can be
missed. These investigators recently used their system to demonstrate
that polymethylmethacrylate (PMMA) lenses are associated with greater
posterior capsule opacification than silicone or acrylic lenses in a
randomized controlled clinical trial.
9 Tetz et
al.
8 described a PCO grading system based on standard
retroillumination photography, in which the observer subjectively
assigns an integer density score from 0 to 4 that is then multiplied by
the fractional area of coverage behind the intraocular lens (IOL) optic
to generate a single metric for PCO. The study was small with only five
eyes evaluated by multiple observers, and one observer re-evaluated
three eyes to assess intraobserver variability. The system was not
assessed for its ability to identify progression.
Pande et al.
10 described a slit lamp–mounted digital
camera for imaging the posterior capsule. The system required special
adaptation of the slit lamp and camera to obtain a coaxial illumination
and imaging path. Ursell et al.
11 used this system to
assess the relationship between intraocular lens material and the
development of PCO. They relied on automated analysis to determine the
presence of texture on the posterior capsule. PCO was present for any
texture above some unstated threshold. The average percent area of the
posterior capsule covered by texture was compared between groups. The
reliability and validity of this system have not been published.
We report the use of a digital camera to obtain retroillumination
photographs of the posterior capsule after cataract extraction and the
analysis of the images with automated algorithms. We evaluated the
validity of this approach, the reproducibility of the image acquisition
and analysis techniques, and the ability of the analysis system to
detect progression in retroillumination photographs, subsequently
digitized, which were taken 2 years apart.
We used a digital retroillumination camera mounted on a slit lamp
to photograph the posterior capsule through a dilated pupil (Marcher
Case 2000 Computerized Anterior Segment Evaluation System, Marcher
Enterprises, Hereford, UK). While seated, the subject is asked to place
his or her chin on a standard slit lamp biomicroscope chin rest and his
or her forehead against a positioning strap. The subject is then asked
to look at a flashing fixation light, and the technician centers the
retroillumination image in the pupil using the computer-generated
centering circles. The camera gain is then adjusted manually before
focusing on the plane of the posterior capsule. The image is obtained
with minimal flash and appears on the screen within 1 to 2 seconds. The
operator has the option to accept or reject the image and can adjust
the gain as needed for the next photograph. The gain adjustments enable
the operator to obtain a wide spread along the gray scale, enhancing
the quality of the image. We required two acceptable images from each
subject, which were archived on the local hard drive. The total time
required was approximately 2 minutes per eye.
We have obtained PCO photographs with this system in approximately 200
patients, none of whom has complained of discomfort from the
illumination system. The reliability and validity data are reported
based on images from this data set. All subjects consented to the
photographs under a research protocol approved by the Johns Hopkins
University School of Medicine Committee on Clinical Investigations.
Three data sets were used to evaluate the validity and
reliability of the image acquisition and analysis system. Face validity
was determined by comparing the clinical grade with that generated by
the automated algorithm using the computer. This was tested on 14 eyes
of 12 subjects graded at slit lamp examination. In addition, a separate
set of 27 digital photographs representing a range of capsular
opacification were graded by consensus by two clinicians (ODS, DSF) and
these grades were compared with the automated analysis. These analyses
were intended to determine whether that which the computer identifies
as PCO is the same thing that a clinician would call PCO, not to
evaluate the reliability of clinicians’ grading of PCO.
Two further analyses were performed to test the reliability of the
system. First, we ran the analytic algorithm on the same set of 20
digital images two times. The results were identical with 100%
correlation, as would be expected using a computer to perform the
analysis. Second, we evaluated test–retest reliability by having the
technician take two sets of digital images of the posterior capsule on
13 subjects 10 minutes apart. Between each set, the subject was
separated from the slit lamp, and the technician left the room, so that
the technician had to reposition the subject before taking the second
set.
We have designed a computer-based image acquisition and grading
scheme that satisfies the needs for a valid and reliable instrument. We
have performed a series of tests to confirm that automated analysis
provides data that are clinically relevant, are reproducible with
little variation, and can be used to detect change over time. The
analysis grades the severity of PCO in a clinically relevant fashion.
The validation studies show that the image analysis system identifies
the percent coverage and density of PCO similarly to subjective
assessment by an ophthalmologist. In addition, clinical grades of PCO
made at the slit lamp examination correlate well with later automated
analysis of retroillumination photographs. The system is ideally suited
for research on the development and prevention of PCO. It was not
designed for use in the clinical setting where the impact of PCO on a
patient’s function may be a more relevant measure.
Others have used the Case 2000 Marcher system to obtain
reproducible images of posterior subcapsular
cataracts.
17 18 We have applied similar strategies to
develop a digital image acquisition system and an automated analysis
system for assessing PCO that yields highly reproducible results. A
second computer analysis of the same images showed no variation. No
human system can perform at this level. Images taken at different times
on the same subject showed minimal variation, especially in determining
percent area covered by PCO. The sources of this variation include
operator variation in image acquisition and image analysis. During the
acquisition stage, several factors can affect reproducibility including
patient fixation, operator focusing techniques, and gain adjustment.
Although the system worked extremely well, we plan to improve
reliability still further by automating the gain-selection process.
During the analysis stage the main source of variability is the
placement of the 4-mm circle that delineates the ROI. The current
approach introduces very little measurement error. However, further
refinement of the system is possible. For example, it is feasible to
establish algorithms that use anatomic landmarks to automate the
placement of the circle in the same place on different photographs
obtained in the same subject.
The Marcher camera has a greater depth of focus than other
retroillumination cameras that have been used to study PCO in the past,
making it more likely that vitreous opacities will be accidentally
interpreted as PCO. Subjective review of the images analyzed using our
system indicated that vitreous opacities contribute, but rarely, to the
computer estimate of the total percent area covered with PCO. However,
analyses of 20 retroillumination images of the posterior capsule taken
within 14 days of cataract extraction found the mean percent covered
was 5.3 ± 3.4 (range, 0.8–11.7). In the three images where
probable vitreous opacities were seen, these opacities covered less
than 0.6% of the ROI. We do not believe that the occasional inclusion
of small vitreous opacities significantly affects the overall image
analysis.
Using an absolute increase in area covered by PCO of 10% or more as a
cutoff, the computer was able to separate progressors from
nonprogressors on digitized photographs taken 2 years apart. This is
especially encouraging, given the significantly worse quality of the
digitized photographs compared with digitally obtained images. Mean
density also increased over the 2-year period, but the difference
identified between the amount of increase in density between
progressors and nonprogressors did not achieve statistical
significance. Average density over the area of interest does not appear
to be an ideal measure of progression, however. When more area is
covered by new, less dense PCO, the average density does not change and
can even decrease. One possible approach to this problem is to use the
SD of the gray scale rather than the mean gray level to calculate
density. A minimally opacified capsule would have a very narrow gray
level distribution and thus a small SD, whereas an opacified capsule
would have a wider range of densities and thus a larger SD. Opacity
could be viewed on a regional basis and algorithms constructed taking
into account local changes throughout the central 4 mm. A third
approach is to assess the texture of the capsule by dividing the SD of
the gray scale by the mean gray level for each distinct area of
opacification. Further research on quantifying these measures will be
helpful in improving the classification scheme for PCO progression.
Image magnification by the cornea can affect the size of the circle
drawn to delineate the ROI. Although this source of variation would not
affect intrasubject assessment of PCO over time, it may theoretically
result in differences in the size of the ROI between subjects, up to
approximately 10%. We are developing algorithms to minimize individual
differences by incorporating keratometry and axial length readings into
the calculation of the ROI.
Some areas of PCO have sharp borders and relatively clear centers. The
system tends to identify the edges of such areas and ignore the clear
central zones. This is a potential limitation of the current algorithms
in detecting the very early stages of capsule opacification. However,
it is not certain that relatively clear central regions are clinically
significant. Newer algorithms that take into account texture may
characterize these clear zones more precisely.
Digital image acquisition is particularly suited to the study of the
biology of posterior capsular opacification. Unlike cataracts,
opacification of the posterior capsule occurs essentially in a single
focal plane. This removes much of the variability and difficulty
associated with trying to photograph the crystalline lens (e.g.,
separating nuclear, cortical, and posterior subcapsular cataracts). In
addition, the technician sees the final image while the subject is
still at the slit lamp. This not only removes the uncertainty about
image quality when taking a photograph, but also allows the technician
to determine directly that the image obtained is consistent with the
clinical findings. This both increases the validity of the data and
avoids the loss of data that can occur when poor photographic quality
precludes accurate grading. Another advantage of digital image
acquisition is enhanced quality control. All images of a particular
type in the data set can be called up with a single command. This rapid
access to the images is impossible using conventional photography.
Finally, the low level of noise in the digital system will enable
clinicians in longitudinal studies to identify changes in posterior
capsular opacity well in advance of the need for laser capsulotomy.
In summary, the digital acquisition and automated analysis of
retroillumination images of the posterior capsule provided reliable and
valid data. The system produced highly reproducible results and drew
conclusions about the capsule similar to those of a human grader who is
grading the capsule both clinically and from a photograph. In addition,
the image analysis system detected progression of capsular opacity in
digitized retroillumination photographs taken 2 years apart.
Reprint requests: Oliver D. Schein, 116 Wilmer, Johns Hopkins Hospital, 600 North Wolfe Street, Baltimore, MD 21287.
Supported by National Eye Institute Grants K08-EY-00358-02 and
R01-EY10857-02, and National Institute on Aging Grant AG-10184. SKW is
a Research to Prevent Blindness senior scientific investigator.
Submitted for publication June 9, 1998; revised January 22, 1999;
accepted February 9, 1999.
Proprietary interest category: N.
The authors thank Stacey Seabrook for her assistance with this
project.
Apple DJ, Solomon KD, Tetz MR, et al. Posterior capsule opacification. Surv Ophthalmol
. 1992;37:73–116.
[CrossRef] [PubMed]Powe NR, Schein OD, Gieser SC, et al. The Cataract Patient Outcome Research Team. Synthesis of the literature on visual acuity and complications following cataract extraction with intraocular lens implantation. Arch Ophthalmol.. 1994;112:239–252.
Javitt JC, Tielsch JM, Canner JK, et al. National outcomes of cataract extraction. Increased risk of retinal complications associated with Nd:YAG laser capsulotomy. Ophthalmology.. 1992;99:1487–1498.
Schein OD, Bass EB, Sharkey P, et al. Cataract surgical techniques: preferences and underlying beliefs. Arch Ophthalmol
. 1995;113:1108–1112.
[CrossRef] [PubMed]Javitt JC, Kendix M, Tielsch JM, et al. Geographic variation in utilization of cataract surgery. Med Care
. 1995;33:90–105.
[CrossRef] [PubMed]Lasa MS, Datiles MB, Magno BV, Mahurkar A. Scheimpflug photography and postcataract surgery posterior capsule opacification. Ophthalmic Surg
. 1995;26:110–113.
[PubMed]Hayashi K, Hayashi H, Nakao F, Hayashi F. In vivo quantitative measurement of posterior capsule opacification after extracapsular cataract surgery. Am J Ophthalmol
. 1998;125:837–843.
[CrossRef] [PubMed]Tetz MR, Auffarth GU, Sperker M, Blum M, Volcker HE. Photographic image analysis system of posterior capsule opacification. J Cataract Refract Surg
. 1997;23:1515–1520.
[CrossRef] [PubMed]Hayashi H, Hayashi K, Nakao F, Hayashi F. Quantitative comparison of posterior capsule opacification after polymethylmethacrylate, silicone, and soft acrylic intraocular lens implantation. Arch Ophthalmol
. 1998;116:1579–1582.
[CrossRef] [PubMed]Pande MV, Ursell PG, Spalton DJ, Heath G, Kundaiker S. High-resolution digital retroillumination imaging of the posterior lens capsule after cataract surgery. J Cataract Refract Surg
. 1997;23:1521–1527.
[CrossRef] [PubMed]Ursell PG, Spalton DJ, Pande MV, et al. Relationship between intraocular lens biomaterials and posterior capsule opacification. J Cataract Refract Surg
. 1998;24:352–360.
[CrossRef] [PubMed]Klette R, Zamperoni P. Handbook of Image Processing Operators. 1996; John Wiley & Sons Chichester, UK.
McCally RL, Farrell RA. Light scattering from cornea and corneal transparency. Masters BR eds. Noninvasive Diagnostic Techniques in Ophthalmology. 1990;189–210. Springer–Verlag New York.
Russ JC. The Image Processing Handbook. 1995; CRC Press Boca Raton, FL.
Paplinski AP, Boyce JF. Segmentation of a class of ophthalmological images using a directional variance operator and co-occurrence arrays. Opt Eng
. 1997;36:3140–3147.
[CrossRef] West SK, Munoz B, Rubin GS, et al. Function and visual impairment in a population-based study of older adults. The SEE project. Salisbury Eye Evaluation. Invest Ophthalmol Vis Sci.. 1997;38:72–82.
Lopez ML, Datiles MB, Podgor MJ, Vivino MA, Mahurkar AA, Lasa SM. Reproducibility study of posterior subcapsular opacities on the NEI retroillumination image analysis system. Eye. 1004;8:657–661.
Sparrow JM, Brown NA, Shun-Shin GA, Bron AJ. The Oxford Modular Cataract Image Analysis System. Eye
. 1990;4:638–648.
[CrossRef] [PubMed]