- Email: [email protected]
Reproducibility of standardized retroillumination photography for quantification of posterior capsule opacification
Reproducibility of standardized retroillumination photography for quantification of posterior capsule opacification Wolf Buehl, MD, Oliver Findl, MD, Rupert Menapace, MD, Michael Georgopoulos, MD, Georg Rainer, MD, Matthias Wirtitsch, MD, Hannes Siegl, MSc, Axel Pinz, PhD ABSTRACT Purpose: To determine the short-term reproducibility of standardized digital retroillumination images of regeneratory posterior capsule opacification (PCO) using the photographic setup at 1 institution. Setting: Department of Ophthalmology, University of Vienna, Vienna, Austria. Methods: In this prospective study, 60 retroillumination images of 30 eyes with varying degrees of PCO and different types of intraocular lenses were acquired with a standardized digital coaxial retroillumination system. Two images were taken per eye with a 1-minute interval between images. Ten other eyes were photographed in the same way but with a 5-day interval between the 2 images. All images were evaluated with a fully automated, objective PCO analysis software in which the PCO score was from 0 to 100. The 2 results (A, B) in each eye were compared, and the differences were calculated. Results: There was a high correlation between the A and B results (r ⫽ 0.99). The mean absolute difference was 3.7%. The repeatability coefficient was 8.8%. Conclusion: Digital coaxial retroillumination photography provided quick acquisition of regeneratory PCO images. It provided excellent image quality and high reproducibility. The technique forms a good basis for automated quantification of PCO with new software systems. J Cataract Refract Surg 2002; 28:265–270 © 2002 ASCRS and ESCRS
P
osterior capsule opacification (PCO) is the most common long-term complication of modern cataract surgery. Clinically, PCO has 2 components: regeneratory and fibrotic. Regeneratory PCO is more common and is the main reason for decreased visual Accepted for publication August 7, 2001. Reprint requests to Oliver Findl, MD, Department of Ophthalmology, University of Vienna, Waehringer Guertel 18-20, 1090 Vienna, Austria. E-mail: [email protected]. © 2002 ASCRS and ESCRS Published by Elsevier Science Inc.
function after intraocular lens (IOL) implantation. It is caused by residual lens epithelial cells (LECs) that migrate and proliferate between the posterior capsule and the IOL, forming monolayers and then Elschnig pearls. This leads to decreased visual acuity and a loss of contrast sensitivity. Posterior capsule opacification can be treated by opening the posterior capsule with a neodymium:YAG (Nd:YAG) laser capsulotomy. However, this may lead to new complications such as cystoid macular edema and 0886-3350/02/$–see front matter PII S0886-3350(01)01228-7
REPRODUCIBILITY OF RETROILLUMINATION FOR PCO QUANTIFICATION
Figure 1. (Buehl) Left: Conventional slitlamp photograph. Right: Coaxial retroillumination image of the same eye. The same digital camera was used for both images.
retinal detachment and does not improve visualization of the peripheral retina. In addition, it increases the cost of cataract treatment. Many researchers have investigated the genesis of PCO, seeking ways to prevent its formation. These efforts include modifications in IOL design and material as well as in surgical technique and the use of drugs. Most clinical trials use subjective scoring techniques or capsulotomy rates as outcome measures.1,2 The former are examiner biased and often time consuming; the latter are dependent on patient demands and financial factors. Therefore, the validity of most trials is questionable and comparison of results among trials is difficult. Thus, the use of a standardized and primarily objective method of PCO quantification is needed. This requires standardized acquisition and automated evaluation systems. Using a slitlamp, regeneratory PCO can be examined best in retroillumination. In contrast, fibrotic PCO can be seen more clearly in reflected light, with a slit angle of about 45 degrees. Thus, standardized retroillumination photography is mandatory for objective analysis of regeneratory PCO with new automated evaluation systems, such as the London POCO3,4 and AQUA software (O. Findl, MD, et al., “AQUA: a Novel Approach to Automated Quantification of After-Cataract,” presented at the XVIIIth Congress of the European Society of Cataract & Refractive Surgeons, Brussels, Belgium, September 2000). 266
Different photographic techniques and assessment systems have been described5– 8 including Scheimpflug photography.9 –12 The Scheimpflug method, however, only assesses the mean opacification of a small part of the posterior capsule. Conventional slitlamp photographs (also in retroillumination) have several disadvantages. They show uneven illumination with loss of detail in the relatively darker area and have a large flash artifact caused by the slit. Therefore, Pande et al.13 developed a retroillumination system with a coaxial optical path that leads to homogeneous illumination over the entire region of interest and causes much smaller light reflections (Figure 1). We adopted this system, digital coaxial retroillumination photography (DCRP), with slight modifications. This method should form a basis for objective analysis of regeneratory PCO using computer programs. To date, little information on the reproducibility of images taken with such a system has been published.14 However, automated quantification of PCO largely depends on the quality of the photographic setup. Even excellent software systems cannot compensate for poor image quality. Thus, the reliability of an image acquisition system should be investigated before new imageanalysis methods are tested. This study evaluated the short-term reproducibility of the standardized digital image-acquisition technique used at our institution using objective PCO analysis software.
J CATARACT REFRACT SURG—VOL 28, FEBRUARY 2002
REPRODUCIBILITY OF RETROILLUMINATION FOR PCO QUANTIFICATION
Patients and Methods Image Acquisition The high-resolution DCRP optical system consists of a Zeiss 30 slitlamp for observation and imaging and a Zeiss retrolux illumination module. Illumination is provided by a Zeiss anterior segment flash pack through a fiber-optic cable. This system allows coaxial illumination and imaging. In a conventional photo slitlamp, the optical path is not coaxial but angled at 14 degrees to the objectives to enable stereopsis. Here, all parts are accurately aligned to facilitate a coaxial optical path to the imaging objective (only 1 of the 2 slitlamp eyepieces is in use). The path continues through a second beam-splitter (45-degree mirror) to the camera. A digital color camera (Kodak NC2000) is connected to the modified system. The main advantage is the higher light sensitivity of the larger, 16.0 mm ⫻ 21.0 mm charge-coupled device (CCD) chip, resulting in a higher signal-to-noise ratio in the acquired images. The CCD has a geometric resolution of 1268 pixels ⫻ 1012 pixels and a 36-bit radiometric resolution (RGB). Via SCSII interface, the images are directly transferred to a personal computer and imported into Adobe Photoshop威 with the Kodak DCS driver plug-in. The best images are saved to a hard disk as tagged image files (TIFF, 3.85 MB per image) for later image analysis. Digital images are obtained with few variations in technique. The patient places his or her chin on the chin rest and is instructed to look straight into the coaxial light. Then, the examiner centers the slitlamp with the slit removed and the posterior capsule is brought into focus. No additional user adjustments are made. The method has some disadvantages. One major drawback is the existence of Purkinje reflections. They are induced by the flashlight, which is reflected at the transitions through the cornea and the IOL. In addition, the optics of the slitlamp cause an internal reflection (Figure 2). Using a fixation light, these reflections can be partly moved out of the region of interest or be overlaid in the center. However, they can not be eliminated entirely. A fixation light was not used in this study because it might have made the system less reproducible. Another problem is artifacts that result from an insufficient tear film. These artifacts may be mistaken for PCO, especially when automated software systems are used for PCO quantification. Therefore, the artifacts must be
Figure 2. (Buehl) In this image P1, P3, and P4 denote Purkinje spots; R indicates the internal reflection caused by the system’s optics.
avoided. In most cases, this can be achieved by rinsing the eye with saline solution before the image is taken. However, patients with severe corneal problems should not be included in clinical trials of PCO because of poor image quality. Study Design and Image Analysis To determine the reproducibility of DCRP, 60 retroillumination photographs of 30 pseudophakic eyes were taken 0 to 3 years after cataract surgery; 2 images were taken per eye. The eyes had received different IOL types and materials and showed varying degrees of PCO ranging from completely clear lenses to severe PCO. All images were taken by the same examiner (W.B.). Between the first and the second image, the patient was instructed to lean back and wait for 1 minute. The position of the photographic device was changed and then readjusted before the second image was taken in the same standardized fashion. A circular region within the capsulorhexis (ie, mostly the central 5.0 mm area but smaller in cases with a smaller capsulorhexis diameter) of each digital image was objectively evaluated using AQUA software. This PC-based application, which assesses PCO using texture analysis with calculation of the entropy of a bitmap, correlates well with subjective PCO scores. The software is fully automated and contains no subjective step in the evaluation process. Any other method that meets these
J CATARACT REFRACT SURG—VOL 28, FEBRUARY 2002
267
REPRODUCIBILITY OF RETROILLUMINATION FOR PCO QUANTIFICATION
Figure 3. (Buehl) Correlation between A and B results (scatterplot).
criteria could be used. The resulting PCO score is between 0 and 100 (0 ⫽ clear posterior capsule; 100 ⫽ very severe PCO). The 2 results (A, B) in each of the 30 eyes were compared, and the differences were calculated. To determine reproducibility, the differences between A and B results were statistically evaluated. As only the short-time reproducibility of DCRP was investigated, longer term reproducibility was assessed by photographing 10 additional eyes on 2 occasions. There was a 5-day interval between the first and the second image. The images were obtained in the same standardized way to examine the reproducibility after a longer period. On both occasions, the eyes were also photographed by a second examiner (examiner 2, M.W.) to determine interobserver variability. Statistical Analysis As all values related to a 0 to 100 scale, they could also be read as percentages. The agreement between repeated measurement by the same operator was specified best by the repeatability coefficient,15 from the standard deviation (approximately 1.96 ⫻ SD). The differences between the results of repeated measurement would not exceed this value with a probability of 95%. In addition, the correlation coefficients (r) and the mean and the maximum difference were calculated. The 95% confidence interval for the mean difference was determined.
Results There was a high correlation between the results of the A and B images (r ⫽ 0.99, P ⬍ .01) (Figure 3), and 268
Figure 4. (Buehl) Top: Distribution of results and absolute differences between A and B results, sorted by the mean result for each image pair. The line connects the lower results. Bottom: Absolute differences between A and B results, sorted by the mean result for each image pair.
there were no significant outliers. The distribution of the results and the absolute values of the 30 differences sorted by the mean AQUA score in each pair of images (Figure 4, top) showed that cases with PCO over the entire range were included in the image data set. The mean absolute difference was 3.7% (confidence interval ⫾0.93%) and the maximum difference, 9.0%. The repeatability coefficient was 8.8%. The absolute values of all differences sorted by the mean result in each case (Figure 4, bottom) showed an equal distribution; thus, the difference between the 2 results in each eye was not dependent on the intensity of the PCO. The analysis of the 10 cases with a 5-day interval between the first and the second image showed similar results. The mean difference between the results of ex-
J CATARACT REFRACT SURG—VOL 28, FEBRUARY 2002
REPRODUCIBILITY OF RETROILLUMINATION FOR PCO QUANTIFICATION
aminer 1 (first image versus second image) was 3.1% ⫾ 2.6% (SD). The maximum difference was 7.7% and the repeatability coefficient, 8.0%. The mean difference between the results of examiner 1 and examiner 2 (first image), a measure of interobserver variability, was 2.5% ⫾ 1.9% (maximum 5.5%) and between examiner 1 (first image) and examiner 2 (second image), 5.0% ⫾ 3.7% (maximum 10.4%). Figure 5 shows 3 examples from the image set. Case 29 had similar results for the A and B images (difference 0.8 on the 0 to 100 scale ⫽ 8 per 1000). Case 3 shows an image pair with a higher deviation (7.2). Case 25 is 1 of the 2 with the highest deviation (8.8).
Discussion Our study found high reproducibility of images taken with the DCRP system. The differences in some cases were to some extent attributable to the diverse flash artifacts and the brightness difference between the 2 images (Figure 5, middle row). Such changes in the intensity of the reflected light are primarily caused by slightly different viewing angles of the patient. It may be possible to solve this problem by using an additional fixation light. Nevertheless, the image pair at the top of Figure 5 led to a minimal difference, even though the location of the flash artifacts is not identical. In contrast, the image pair at the bottom of Figure 5 produced a higher difference; however, there seems to be little difference between the 2 images at first view. Image B is, however, slightly out of focus. This suggests that both image definition and image brilliance had an impact on the evaluation using the analysis software. In addition, the evaluated area is not exactly the same in some image pairs, as in this one, because of a slight variation in the viewing angle. Barman et al.3 found a repeatability of 9.8% based on a data set of retroillumination images of 32 eyes. The images were taken 1 week apart and were evaluated with the POCO software. These results are comparable to our findings. In our study, we examined mainly the shorttime reproducibility of the DCRP technique. The results in the 10 eyes that were photographed on 2 separate occasions were almost identical to the results in the previous 30 cases. This shows that the image acquisition system is reproducible, relatively objective, and almost operator independent.
Figure 5. (Buehl) Image pair with small (top), larger (middle), and maximum (bottom) difference in results.
Camparini and coauthors8 state that reflected-light photography (slit angle 45 degrees) is a better method of capturing the severity of PCO and that the use of retroillumination images may lead to its underestimation. However, they used a standard slitlamp and did not differentiate between regeneratory and fibrotic PCO. The fibrotic component of PCO cannot be visualized satisfactorily in retroilluminated images. On the other hand, it makes sense to distinguish between the 2 components. Fibrotic PCO can be shown adequately in reflected-light images. Coaxial retroillumination photography is the best way to capture regeneratory PCO. Variations in background illumination and the Purkinje reflections remain significant problems in the retroillumination system we used. It is possible to eliminate the flash artifacts in the digital images by recording a series of images of the same eye (in different directions of
J CATARACT REFRACT SURG—VOL 28, FEBRUARY 2002
269
REPRODUCIBILITY OF RETROILLUMINATION FOR PCO QUANTIFICATION
view) and fusing several images as the Purkinje spots change position with varying direction of gaze. Three images of each eye taken sequentially in 3 different directions of gaze should remove all reflections. By fusing valid information (areas without reflections) taken from the 3 images (1 reference, 2 additional), a combined image free of reflections can be rendered. To record the image sets (3 images per eye) in a standardized manner, we added a ring with 3 red fixation light-emitting diodes (LEDs) to the image acquisition system in front of the slitlamp. Three images are taken with the patient looking straight, up, and temporal, the latter 2 by fixating on the corresponding LED. The LEDs can be switched on and off by the examiner. The active LED goes off automatically when acquiring the image to avoid additional reflections by the red light. However, several problems remain to be solved in the development of an effective algorithm for image fusion. Objective and automated quantification of PCO is gaining in importance, but it is dependent on highquality image acquisition. The DCRP method is a simple, quick way to document regeneratory PCO. Although some problems remain, DCRP offers good image quality and shows high reproducibility, with a repeatability coefficient of 8.8%. Therefore, DCRP provides an excellent basis for automated assessment of regeneratory PCO.
7.
8.
9.
10.
11.
12.
13.
14.
References 1. Schaumberg DA, Dana MR, Christen WG, Glynn RJ. A systematic overview of the incidence of posterior capsule opacification. Ophthalmology 1998; 105:1213–1221 2. Apple DJ, Solomon KD, Tetz MR, et al. Posterior capsule opacification. Surv Ophthalmol 1992; 37:73–116 3. Barman SA, Hollick EJ, Boyce JF, et al. Quantification of posterior capsular opacification in digital images after cataract surgery. Invest Ophthalmol Vis Sci 2000; 41: 3882–3892 4. 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 5. Tetz MR, Auffarth GU, Sperker M, et al. Photographic image analysis system of posterior capsule opacification. J Cataract Refract Surg 1997; 23:1515–1520 6. Wang M-C, Woung L-C. Digital retroilluminated photography to analyze posterior capsule opacification in
270
15.
eyes with intraocular lenses. J Cataract Refract Surg 2000; 26:56 – 61 Friedman DS, Duncan DD, Munoz B, et al. Digital image capture and automated analysis of posterior capsular opacification. Invest Ophthalmol Vis Sci 1999; 40:1715– 1726 Camparini M, Macaluso C, Reggiani L, Maraini G. Retroillumination versus reflected-light images in the photographic assessment of posterior capsule opacification. Invest Ophthalmol Vis Sci 2000; 41:3074 –3079 Lasa MSM, Datiles MB III, Magno BV, Mahurkar A. Scheimpflug photography and postcataract surgery posterior capsule opacification. Ophthalmic Surg 1995; 26: 110 –113 Klos KM, Richter R, Schnaudigel O-E, Ohrloff C. Image analysis of implanted rigid and foldable intraocular lenses in human eyes using Scheimpflug photography. Ophthalmic Res 1999; 31:130 –133 Hayashi K, Hayashi H, Nakao F, Hayashi F. Reproducibility of posterior capsule opacification measurement using Scheimpflug videophotography. J Cataract Refract Surg 1998; 24:1632–1635 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 Pande MV, Ursell PG, Spalton DJ, et al. High-resolution digital retroillumination imaging of the posterior lens capsule after cataract surgery. J Cataract Refract Surg 1997; 23:1521–1527 Hollick EJ, Spalton DJ, Ursell PG, et al. The effect of polymethylmethacrylate, silicone, and polyacrylic intraocular lenses on posterior capsular opacification 3 years after cataract surgery. Ophthalmology 1999; 106: 49 –54; discussion by RC Drews, 54 –55 Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1:307–310
From the Department of Ophthalmology, University of Vienna, Vienna, (Buehl, Findl, Menapace, Georgopoulos, Rainer, Wirtitsch), and Institute of Electrical Measurement and Measurement Signal Processing, Graz Technical University Graz (Siegl, Pinz), Austria. Presented in part at the 15th Congress of the German Society for IOL Implantation, Bremen, Germany, February 2001. Funded in part by grant P14269-MED from Fonds zur Fo¨rderung der wissenschaftlichen Forschung, Vienna, Austria. None of the authors has a financial or proprietary interest in any material or method mentioned. Dr. David Spalton, St. Thomas Hospital, and Gerald Heath, Zeiss, London, England, and Hans Foerster, Zeiss, Vienna, Austria, helped assemble the photographic setup.
J CATARACT REFRACT SURG—VOL 28, FEBRUARY 2002
P
osterior capsule opacification (PCO) is the most common long-term complication of modern cataract surgery. Clinically, PCO has 2 components: regeneratory and fibrotic. Regeneratory PCO is more common and is the main reason for decreased visual Accepted for publication August 7, 2001. Reprint requests to Oliver Findl, MD, Department of Ophthalmology, University of Vienna, Waehringer Guertel 18-20, 1090 Vienna, Austria. E-mail: [email protected]. © 2002 ASCRS and ESCRS Published by Elsevier Science Inc.
function after intraocular lens (IOL) implantation. It is caused by residual lens epithelial cells (LECs) that migrate and proliferate between the posterior capsule and the IOL, forming monolayers and then Elschnig pearls. This leads to decreased visual acuity and a loss of contrast sensitivity. Posterior capsule opacification can be treated by opening the posterior capsule with a neodymium:YAG (Nd:YAG) laser capsulotomy. However, this may lead to new complications such as cystoid macular edema and 0886-3350/02/$–see front matter PII S0886-3350(01)01228-7
REPRODUCIBILITY OF RETROILLUMINATION FOR PCO QUANTIFICATION
Figure 1. (Buehl) Left: Conventional slitlamp photograph. Right: Coaxial retroillumination image of the same eye. The same digital camera was used for both images.
retinal detachment and does not improve visualization of the peripheral retina. In addition, it increases the cost of cataract treatment. Many researchers have investigated the genesis of PCO, seeking ways to prevent its formation. These efforts include modifications in IOL design and material as well as in surgical technique and the use of drugs. Most clinical trials use subjective scoring techniques or capsulotomy rates as outcome measures.1,2 The former are examiner biased and often time consuming; the latter are dependent on patient demands and financial factors. Therefore, the validity of most trials is questionable and comparison of results among trials is difficult. Thus, the use of a standardized and primarily objective method of PCO quantification is needed. This requires standardized acquisition and automated evaluation systems. Using a slitlamp, regeneratory PCO can be examined best in retroillumination. In contrast, fibrotic PCO can be seen more clearly in reflected light, with a slit angle of about 45 degrees. Thus, standardized retroillumination photography is mandatory for objective analysis of regeneratory PCO with new automated evaluation systems, such as the London POCO3,4 and AQUA software (O. Findl, MD, et al., “AQUA: a Novel Approach to Automated Quantification of After-Cataract,” presented at the XVIIIth Congress of the European Society of Cataract & Refractive Surgeons, Brussels, Belgium, September 2000). 266
Different photographic techniques and assessment systems have been described5– 8 including Scheimpflug photography.9 –12 The Scheimpflug method, however, only assesses the mean opacification of a small part of the posterior capsule. Conventional slitlamp photographs (also in retroillumination) have several disadvantages. They show uneven illumination with loss of detail in the relatively darker area and have a large flash artifact caused by the slit. Therefore, Pande et al.13 developed a retroillumination system with a coaxial optical path that leads to homogeneous illumination over the entire region of interest and causes much smaller light reflections (Figure 1). We adopted this system, digital coaxial retroillumination photography (DCRP), with slight modifications. This method should form a basis for objective analysis of regeneratory PCO using computer programs. To date, little information on the reproducibility of images taken with such a system has been published.14 However, automated quantification of PCO largely depends on the quality of the photographic setup. Even excellent software systems cannot compensate for poor image quality. Thus, the reliability of an image acquisition system should be investigated before new imageanalysis methods are tested. This study evaluated the short-term reproducibility of the standardized digital image-acquisition technique used at our institution using objective PCO analysis software.
J CATARACT REFRACT SURG—VOL 28, FEBRUARY 2002
REPRODUCIBILITY OF RETROILLUMINATION FOR PCO QUANTIFICATION
Patients and Methods Image Acquisition The high-resolution DCRP optical system consists of a Zeiss 30 slitlamp for observation and imaging and a Zeiss retrolux illumination module. Illumination is provided by a Zeiss anterior segment flash pack through a fiber-optic cable. This system allows coaxial illumination and imaging. In a conventional photo slitlamp, the optical path is not coaxial but angled at 14 degrees to the objectives to enable stereopsis. Here, all parts are accurately aligned to facilitate a coaxial optical path to the imaging objective (only 1 of the 2 slitlamp eyepieces is in use). The path continues through a second beam-splitter (45-degree mirror) to the camera. A digital color camera (Kodak NC2000) is connected to the modified system. The main advantage is the higher light sensitivity of the larger, 16.0 mm ⫻ 21.0 mm charge-coupled device (CCD) chip, resulting in a higher signal-to-noise ratio in the acquired images. The CCD has a geometric resolution of 1268 pixels ⫻ 1012 pixels and a 36-bit radiometric resolution (RGB). Via SCSII interface, the images are directly transferred to a personal computer and imported into Adobe Photoshop威 with the Kodak DCS driver plug-in. The best images are saved to a hard disk as tagged image files (TIFF, 3.85 MB per image) for later image analysis. Digital images are obtained with few variations in technique. The patient places his or her chin on the chin rest and is instructed to look straight into the coaxial light. Then, the examiner centers the slitlamp with the slit removed and the posterior capsule is brought into focus. No additional user adjustments are made. The method has some disadvantages. One major drawback is the existence of Purkinje reflections. They are induced by the flashlight, which is reflected at the transitions through the cornea and the IOL. In addition, the optics of the slitlamp cause an internal reflection (Figure 2). Using a fixation light, these reflections can be partly moved out of the region of interest or be overlaid in the center. However, they can not be eliminated entirely. A fixation light was not used in this study because it might have made the system less reproducible. Another problem is artifacts that result from an insufficient tear film. These artifacts may be mistaken for PCO, especially when automated software systems are used for PCO quantification. Therefore, the artifacts must be
Figure 2. (Buehl) In this image P1, P3, and P4 denote Purkinje spots; R indicates the internal reflection caused by the system’s optics.
avoided. In most cases, this can be achieved by rinsing the eye with saline solution before the image is taken. However, patients with severe corneal problems should not be included in clinical trials of PCO because of poor image quality. Study Design and Image Analysis To determine the reproducibility of DCRP, 60 retroillumination photographs of 30 pseudophakic eyes were taken 0 to 3 years after cataract surgery; 2 images were taken per eye. The eyes had received different IOL types and materials and showed varying degrees of PCO ranging from completely clear lenses to severe PCO. All images were taken by the same examiner (W.B.). Between the first and the second image, the patient was instructed to lean back and wait for 1 minute. The position of the photographic device was changed and then readjusted before the second image was taken in the same standardized fashion. A circular region within the capsulorhexis (ie, mostly the central 5.0 mm area but smaller in cases with a smaller capsulorhexis diameter) of each digital image was objectively evaluated using AQUA software. This PC-based application, which assesses PCO using texture analysis with calculation of the entropy of a bitmap, correlates well with subjective PCO scores. The software is fully automated and contains no subjective step in the evaluation process. Any other method that meets these
J CATARACT REFRACT SURG—VOL 28, FEBRUARY 2002
267
REPRODUCIBILITY OF RETROILLUMINATION FOR PCO QUANTIFICATION
Figure 3. (Buehl) Correlation between A and B results (scatterplot).
criteria could be used. The resulting PCO score is between 0 and 100 (0 ⫽ clear posterior capsule; 100 ⫽ very severe PCO). The 2 results (A, B) in each of the 30 eyes were compared, and the differences were calculated. To determine reproducibility, the differences between A and B results were statistically evaluated. As only the short-time reproducibility of DCRP was investigated, longer term reproducibility was assessed by photographing 10 additional eyes on 2 occasions. There was a 5-day interval between the first and the second image. The images were obtained in the same standardized way to examine the reproducibility after a longer period. On both occasions, the eyes were also photographed by a second examiner (examiner 2, M.W.) to determine interobserver variability. Statistical Analysis As all values related to a 0 to 100 scale, they could also be read as percentages. The agreement between repeated measurement by the same operator was specified best by the repeatability coefficient,15 from the standard deviation (approximately 1.96 ⫻ SD). The differences between the results of repeated measurement would not exceed this value with a probability of 95%. In addition, the correlation coefficients (r) and the mean and the maximum difference were calculated. The 95% confidence interval for the mean difference was determined.
Results There was a high correlation between the results of the A and B images (r ⫽ 0.99, P ⬍ .01) (Figure 3), and 268
Figure 4. (Buehl) Top: Distribution of results and absolute differences between A and B results, sorted by the mean result for each image pair. The line connects the lower results. Bottom: Absolute differences between A and B results, sorted by the mean result for each image pair.
there were no significant outliers. The distribution of the results and the absolute values of the 30 differences sorted by the mean AQUA score in each pair of images (Figure 4, top) showed that cases with PCO over the entire range were included in the image data set. The mean absolute difference was 3.7% (confidence interval ⫾0.93%) and the maximum difference, 9.0%. The repeatability coefficient was 8.8%. The absolute values of all differences sorted by the mean result in each case (Figure 4, bottom) showed an equal distribution; thus, the difference between the 2 results in each eye was not dependent on the intensity of the PCO. The analysis of the 10 cases with a 5-day interval between the first and the second image showed similar results. The mean difference between the results of ex-
J CATARACT REFRACT SURG—VOL 28, FEBRUARY 2002
REPRODUCIBILITY OF RETROILLUMINATION FOR PCO QUANTIFICATION
aminer 1 (first image versus second image) was 3.1% ⫾ 2.6% (SD). The maximum difference was 7.7% and the repeatability coefficient, 8.0%. The mean difference between the results of examiner 1 and examiner 2 (first image), a measure of interobserver variability, was 2.5% ⫾ 1.9% (maximum 5.5%) and between examiner 1 (first image) and examiner 2 (second image), 5.0% ⫾ 3.7% (maximum 10.4%). Figure 5 shows 3 examples from the image set. Case 29 had similar results for the A and B images (difference 0.8 on the 0 to 100 scale ⫽ 8 per 1000). Case 3 shows an image pair with a higher deviation (7.2). Case 25 is 1 of the 2 with the highest deviation (8.8).
Discussion Our study found high reproducibility of images taken with the DCRP system. The differences in some cases were to some extent attributable to the diverse flash artifacts and the brightness difference between the 2 images (Figure 5, middle row). Such changes in the intensity of the reflected light are primarily caused by slightly different viewing angles of the patient. It may be possible to solve this problem by using an additional fixation light. Nevertheless, the image pair at the top of Figure 5 led to a minimal difference, even though the location of the flash artifacts is not identical. In contrast, the image pair at the bottom of Figure 5 produced a higher difference; however, there seems to be little difference between the 2 images at first view. Image B is, however, slightly out of focus. This suggests that both image definition and image brilliance had an impact on the evaluation using the analysis software. In addition, the evaluated area is not exactly the same in some image pairs, as in this one, because of a slight variation in the viewing angle. Barman et al.3 found a repeatability of 9.8% based on a data set of retroillumination images of 32 eyes. The images were taken 1 week apart and were evaluated with the POCO software. These results are comparable to our findings. In our study, we examined mainly the shorttime reproducibility of the DCRP technique. The results in the 10 eyes that were photographed on 2 separate occasions were almost identical to the results in the previous 30 cases. This shows that the image acquisition system is reproducible, relatively objective, and almost operator independent.
Figure 5. (Buehl) Image pair with small (top), larger (middle), and maximum (bottom) difference in results.
Camparini and coauthors8 state that reflected-light photography (slit angle 45 degrees) is a better method of capturing the severity of PCO and that the use of retroillumination images may lead to its underestimation. However, they used a standard slitlamp and did not differentiate between regeneratory and fibrotic PCO. The fibrotic component of PCO cannot be visualized satisfactorily in retroilluminated images. On the other hand, it makes sense to distinguish between the 2 components. Fibrotic PCO can be shown adequately in reflected-light images. Coaxial retroillumination photography is the best way to capture regeneratory PCO. Variations in background illumination and the Purkinje reflections remain significant problems in the retroillumination system we used. It is possible to eliminate the flash artifacts in the digital images by recording a series of images of the same eye (in different directions of
J CATARACT REFRACT SURG—VOL 28, FEBRUARY 2002
269
REPRODUCIBILITY OF RETROILLUMINATION FOR PCO QUANTIFICATION
view) and fusing several images as the Purkinje spots change position with varying direction of gaze. Three images of each eye taken sequentially in 3 different directions of gaze should remove all reflections. By fusing valid information (areas without reflections) taken from the 3 images (1 reference, 2 additional), a combined image free of reflections can be rendered. To record the image sets (3 images per eye) in a standardized manner, we added a ring with 3 red fixation light-emitting diodes (LEDs) to the image acquisition system in front of the slitlamp. Three images are taken with the patient looking straight, up, and temporal, the latter 2 by fixating on the corresponding LED. The LEDs can be switched on and off by the examiner. The active LED goes off automatically when acquiring the image to avoid additional reflections by the red light. However, several problems remain to be solved in the development of an effective algorithm for image fusion. Objective and automated quantification of PCO is gaining in importance, but it is dependent on highquality image acquisition. The DCRP method is a simple, quick way to document regeneratory PCO. Although some problems remain, DCRP offers good image quality and shows high reproducibility, with a repeatability coefficient of 8.8%. Therefore, DCRP provides an excellent basis for automated assessment of regeneratory PCO.
7.
8.
9.
10.
11.
12.
13.
14.
References 1. Schaumberg DA, Dana MR, Christen WG, Glynn RJ. A systematic overview of the incidence of posterior capsule opacification. Ophthalmology 1998; 105:1213–1221 2. Apple DJ, Solomon KD, Tetz MR, et al. Posterior capsule opacification. Surv Ophthalmol 1992; 37:73–116 3. Barman SA, Hollick EJ, Boyce JF, et al. Quantification of posterior capsular opacification in digital images after cataract surgery. Invest Ophthalmol Vis Sci 2000; 41: 3882–3892 4. 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 5. Tetz MR, Auffarth GU, Sperker M, et al. Photographic image analysis system of posterior capsule opacification. J Cataract Refract Surg 1997; 23:1515–1520 6. Wang M-C, Woung L-C. Digital retroilluminated photography to analyze posterior capsule opacification in
270
15.
eyes with intraocular lenses. J Cataract Refract Surg 2000; 26:56 – 61 Friedman DS, Duncan DD, Munoz B, et al. Digital image capture and automated analysis of posterior capsular opacification. Invest Ophthalmol Vis Sci 1999; 40:1715– 1726 Camparini M, Macaluso C, Reggiani L, Maraini G. Retroillumination versus reflected-light images in the photographic assessment of posterior capsule opacification. Invest Ophthalmol Vis Sci 2000; 41:3074 –3079 Lasa MSM, Datiles MB III, Magno BV, Mahurkar A. Scheimpflug photography and postcataract surgery posterior capsule opacification. Ophthalmic Surg 1995; 26: 110 –113 Klos KM, Richter R, Schnaudigel O-E, Ohrloff C. Image analysis of implanted rigid and foldable intraocular lenses in human eyes using Scheimpflug photography. Ophthalmic Res 1999; 31:130 –133 Hayashi K, Hayashi H, Nakao F, Hayashi F. Reproducibility of posterior capsule opacification measurement using Scheimpflug videophotography. J Cataract Refract Surg 1998; 24:1632–1635 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 Pande MV, Ursell PG, Spalton DJ, et al. High-resolution digital retroillumination imaging of the posterior lens capsule after cataract surgery. J Cataract Refract Surg 1997; 23:1521–1527 Hollick EJ, Spalton DJ, Ursell PG, et al. The effect of polymethylmethacrylate, silicone, and polyacrylic intraocular lenses on posterior capsular opacification 3 years after cataract surgery. Ophthalmology 1999; 106: 49 –54; discussion by RC Drews, 54 –55 Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1:307–310
From the Department of Ophthalmology, University of Vienna, Vienna, (Buehl, Findl, Menapace, Georgopoulos, Rainer, Wirtitsch), and Institute of Electrical Measurement and Measurement Signal Processing, Graz Technical University Graz (Siegl, Pinz), Austria. Presented in part at the 15th Congress of the German Society for IOL Implantation, Bremen, Germany, February 2001. Funded in part by grant P14269-MED from Fonds zur Fo¨rderung der wissenschaftlichen Forschung, Vienna, Austria. None of the authors has a financial or proprietary interest in any material or method mentioned. Dr. David Spalton, St. Thomas Hospital, and Gerald Heath, Zeiss, London, England, and Hans Foerster, Zeiss, Vienna, Austria, helped assemble the photographic setup.
J CATARACT REFRACT SURG—VOL 28, FEBRUARY 2002