CLINICAL SCIENCE
Comparison of 4 Specular Microscopes in Healthy Eyes and Eyes With Cornea Guttata or Corneal Grafts Nikolaus Luft, MD,* Nino Hirnschall, MD, PhD,* Sandra Schuschitz, MSc,* Petra Draschl, MD,* and Oliver Findl, MD, MBA*†
Purpose: The aim of this study was to compare 4 noncontact specular microscopes for the assessment of the corneal endothelium in a heterogeneous sample of eyes.
Methods: This prospective study included 22 healthy eyes, 16 eyes with cornea guttata and 31 eyes that had undergone penetrating keratoplasty or Descemet stripping automated endothelial keratoplasty. Corneal endothelial cell parameters of all eyes were assessed with the CellChek XL (Konan Medical, Hyogo, Japan), Perseus (Bon Optic, Lübeck, Germany), EM-3000 (Tomey, Nagoya, Japan), and CEM-530 (Nidek Co, Ltd, Gamagori, Japan) in a randomized order. Bland–Altman plots of interdevice agreement were produced. The time taken for data entry, scanning, and automated image analysis was recorded, and instrument repeatability was analyzed. Results: The mean age was 61 6 19 years (range, 24–88). The measurement success rate was 100% in healthy corneas, ranged between 64.5% and 93.5% in eyes with corneal grafts, and between 0% and 18.8% in eyes with cornea guttata. Comparative analysis of endothelial cell parameters revealed significant differences in endothelial cell density readings in the total study population and in subgroups (P , 0.01). Conclusions: The CEM-530 and Perseus were extremely fast, showed highly repeatable measurements, and are recommendable for screening purposes. The CellChek XL device was slower, but extensive postprocessing options and an exportable database make this instrument suitable for compromised corneas and for research purposes. The EM-3000 device ranged between the fast devices and the CellChek XL and comes with reliable automated cell-counting software. However, careful interpretation of results from automated image analysis software is mandatory. Key Words: specular microscopy, endothelial cell density, dystrophy, guttata (Cornea 2015;34:381–386)
Received for publication August 15, 2014; revision received December 28, 2014; accepted December 29, 2014. Published online ahead of print February 19, 2015. From the *VIROS—Vienna Institute for Research in Ocular Surgery, A Karl Landsteiner Institute, Hanusch Hospital, Vienna, Austria; and †Moorfields Eye Hospital NHS Foundation Trust, London, United Kingdom. The authors have no funding or conflicts of interest to disclose. Reprints: Oliver Findl, MD, MBA, Hanusch Hospital, Heinrich-CollinStrasse 30, 1140 Vienna, Austria (e-mail: oliver@findl.at). Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Cornea Volume 34, Number 4, April 2015
B
ecause of the inability of corneal endothelial cells to regenerate, it is paramount to detect patients with endothelial cell dystrophies or low levels of endothelial cells in routine preoperative testing, such as before cataract surgery.1 Moreover, reliable assessment of corneal endothelial health is mandatory in surveillance of patients with phakic intraocular lenses.2 A further useful application for corneal endothelial cell measurements is studies evaluating the effect of surgical techniques or devices on the corneal endothelium.3,4 Specular microscopes allow a quantitative and morphological assessment of corneal endothelial cells. The purpose of this study was to compare 4 commercially available, noncontact specular microscopes in a heterogeneous sample of eyes.
MATERIALS AND METHODS This prospective study included patients who had undergone Descemet stripping automated endothelial keratoplasty (DSAEK) or penetrating keratoplasty (PKP), patients with cornea guttata or Fuchs endothelial dystrophy (FED), and subjects with healthy corneas. Eyes with severe corneal scarring were excluded from this study. All the research and measurements followed the tenets of the Declaration of Helsinki, and all participants gave informed consent before the start of the study. Three examiners who were inexperienced in using all 4 investigated noncontact specular microscope systems, namely CellChek XL (Konan Medical, Hyogo, Japan), Perseus (Bon Optic, Lübeck, Germany), EM-3000 (Tomey, Nagoya, Japan), and CEM-530 (Nidek Co, Ltd, Gamagori, Japan), received an introduction for each of the 4 devices by experienced examiners. On the next day, both eyes of each subject were measured with the 4 specular microscopes in a single office visit. The instruments were used in a randomized order and 3 consecutive measurements of each eye were performed. Each patient was randomly assigned to 1 of the 3 examiners who then performed all measurements with all devices on that particular patient. After each scan, the patient was asked to lean back and the chinrest was readjusted. In all cases, the patient fixated on a central fixation light. In case of an acquired endothelial image without visible, well-defined cells, the scan was classified as unsuccessful. The following parameters were recorded for each eye: endothelial cell density (ECD) (cells/mm2), coefficient of variation of the cell area (CV), cell hexagonality (proportion of hexagonal cells in percentage), central corneal thickness www.corneajrnl.com |
381
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Luft et al
(mm), and the number of cells analyzed. Furthermore, the time taken for data entry, scanning, and automated image analysis was recorded for one randomly elected eye of each patient using a digital stopwatch. The timed examination was the first of 3 examinations on the respective eye in all cases. Additionally, a comprehensive anterior segment slit-lamp examination of all subjects was conducted at the beginning of their visit.
CellChek XL The CellChek XL specular microscope system captures a photographic field of 0.1 mm2. After subsequent autoalignment and automated image acquisition, image analysis was performed using the built-in automated cell-counting software in default settings. For autoanalysis, the software uses predefined reference patterns with different cell sizes (“Small,” “Medium,” “Large,” and “XLarge”) and defaults to “Small.” The autoanalyzed image was saved electronically to a database implemented in the CellChek XL system. Additionally, manual (“semiautomated”) evaluation of all endothelium photographs was performed. Different editing tools of the onboard software were used as advised in the instruction manual: selecting an appropriate cell size reference pattern, resizing the area to be analyzed by the autodetection algorithm, and the “Center method” or “Flex-Center method”5 for manual cell counting.
Perseus With the Perseus specular microscope, up to 300 endothelial cells can be counted within an area of 0.2 mm2. The Perseus provides a touch screen to roughly align the eye to be examined. After subsequent automated alignment and autoacquisition, the captured image was analyzed by the cellidentification software and saved to the onboard database. In case of images with visible cells that were missed by the cellidentification algorithm, the manual analysis method for identifying endothelial was performed instead in accordance with the instruction manual: cell corner tracing with a touch pen, erasing incorrectly traced cell borders, excluding areas from analysis that are obscured (eg, by guttae).
EM-3000 The EM-3000 specular microscope uses an optical magnification of ·190 and up to 300 cells per image are counted within an area of 0.1 mm2. This device automatically captured 15 images per measurement, which were displayed and sorted by the software with respect to image quality. The image with highest quality was then selected by the examiner and automated cell detection, and counting was performed using the built-in manufacturer’s software. Manual image analysis was not performed with the EM-3000 because the study device was not supplied with the proprietary database that allows manual image after processing.
Cornea Volume 34, Number 4, April 2015
captures a total of 16 endothelial photographs per scan, which are displayed and automatically sorted based on quality immediately after scanning. The most suitable image according to the judgment of the examiner was elected for automated cell detection with the manufacturer’s software. Because the CEM-530 offered only a very basic method for manual cell detection, we abstained from performing manual analysis with this machine.
Statistical Analysis All statistical analysis was performed using Excel Version 14.4.1 for Mac (Microsoft Corporation) and SPSS Statistics Version 21.0.0.0 for Mac (IBM). The Kolmogorov–Smirnov test was used to confirm the normality of data distribution. Descriptive data are always presented as mean, SD, and range. P , 0.05 was considered statistically significant. A Fisher exact test with a Bonferroni correction was used for multiple comparisons of measurement success rate between devices (nonmetric data). For comparative statistical analysis of endothelial cell parameters, the median value of 3 consecutive measurements of 1 device was taken into account. In the case of 2 successful scans, the mean value was used instead and if only 1 measurement was successful, this measurement was used for further analysis. A repeatedmeasures analysis of variance (ANOVA) with a Bonferroni correction was used to compare ECD, CV, cell hexagonality, and central corneal thickness readings between instruments and to determine whether these differed significantly. Semiautomated analysis with the CellChek XL was defined as the gold standard for quantitative and qualitative assessment of the endothelial cells. Agreement of ECD readings between the respective devices was evaluated using the Bland–Altman method.6 The limits of agreement (LoA) were defined as the mean interdevice difference 6 2 SD. For repeatability testing, a 1-way ANOVA was applied to compare mean SD of 3 consecutive ECD readings between devices. A Friedman test and post hoc analysis with Wilcoxon signed-rank test conducted with a Bonferroni correction were performed to assess differences in time measurements between the instruments.
RESULTS
The CEM-530 noncontact specular microscope captures an endothelial image 0.1 mm2 in size. This device
The study included 69 eyes of 36 patients: 22 healthy eyes, 16 eyes with FED, and 31 eyes with corneal grafts (29 DSAEK eyes and 2 PKP eyes). Male to female ratio was 15 (42%) to 21 (58%). The mean age was 61 6 19 years (range, 24–88). Three eyes were excluded from the study: the first was an eye that had undergone PKP with a decompensated cornea and phthisis. The second had end-stage corneal dystrophy, severe stromal vascularization, and calcification. In the third case, the subject withdrew from the trial for personal reasons before measurements with all 4 devices could be completed. For semiautomated analysis of the acquired images with the CellChek XL, cell size patterns other than the default “small” pattern were selected in 36 scans of 16 eyes. In 29 scans of 15 eyes, the area of analysis in the endothelial
382
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
CEM-530
| www.corneajrnl.com
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Cornea Volume 34, Number 4, April 2015
Comparison of 4 Specular Microscopes
photograph was cropped. Moreover, the “Center method” was applied in 29 scans of 12 eyes, and the “Flex-center” method was used in 12 scans of 6 eyes. With these means of manual image analysis, no additional scans were obtainable compared with automated analysis. With the Perseus, manual image analysis was performed on 5 endothelial photographs of 2 eyes by tracing the cell borders. In neither of these 2 eyes, successful scans had been obtainable with automated image analysis. A summary of successful endothelial cell measurements with the 4 instruments is given in Table 1. No significant difference in the success rate of measurements between instruments was observed (all with P . 0.05).
Comparative Analysis
and the CEM-530 were significantly higher compared with the gold standard with mean differences of 681, 166, and 262 cells per square millimeter, respectively (all with P , 0.01). In DSAEK eyes, only the EM-3000 produced ECD readings that were comparable with the gold standard device (mean difference, 119 cells/mm2; P = 0.31).
Repeatability The mean SD of 3 consecutive ECD measurements with each instrument are shown in Table 2. A 1-way ANOVA revealed no statistically significant differences in repeatability between the 4 instruments (P = 0.304).
Time Measurements
A minimum of 1 image suitable for automated analysis was obtainable with all 4 instruments in 39 eyes of 26 patients (22 healthy eyes and 17 DSAEK eyes). With the Perseus, manual image analysis was performed in none of these eyes, and thus, no differentiation between manual and automated analysis was made for this instrument in comparative analysis. The quantitative and morphological endothelial cell parameters measured with each device are summarized in Table 2. Mean ECD measurements obtained with the Perseus, the EM-3000, or the CEM-530 did not differ significantly compared with those obtained with the gold standard device with mean differences of 6, 26, and 4 cells per square millimeter, respectively (all with P = 0.99). In contrast, automated image analysis with the CellChek XL produced ECD readings that were significantly higher (mean difference, 391 cells/mm2; P , 0.01). Bland–Altman plots of interdevice differences (Fig. 1) also demonstrate that the difference in ECD readings was largest between automated image analysis with the CellChek XL and the gold standard. In the subgroup of healthy corneas, no significant differences in ECD readings were detected between the Perseus, the EM-3000, and the CEM-530. However, these 3 instruments understated ECD compared with the gold standard with mean differences of 117, 137, and 196 cells per square millimeter, respectively (all with P , 0.01). Automated analysis with the CellChek XL produced readings comparable with the reference device with a mean difference of 167 cells per square millimeter (P = 0.99). However, ECD readings with these 2 methods only differed in 3 cases (Fig. 1A). In the DSAEK cohort, ECD readings obtained with the CellChek XL using automated image analysis, the Perseus,
The mean total measurement time was 72 6 20 seconds (range, 46–121) with the CellChek XL, 54 6 21 seconds (range, 36–160) with the Perseus, 64 6 19 seconds (range, 36– 144) with the EM-3000, and 53 6 12 seconds (range, 37–91) with the CEM-530. The Friedman test and a post hoc Wilcoxon signed-rank test conducted with a Bonferroni correction applied (resulting in a significance level set at P , 0.0083) revealed no significant difference in mean total measurement time between the Perseus and the CEM-530 (P = 0.75). Equally, no difference was observed between the CellChek XL and the EM-3000 (P = 0.039). Measurements with the latter devices took significantly longer than with the former 2 (all with P , 0.003). Box plots illustrating the mean times taken for data entry, scanning, and automated image analysis with the 4 instruments are shown in Figure 2. Significant differences between devices were discovered in the mean times taken for patient data entry and for automated image analysis (P , 0.001). No significant difference was found in the mean times taken for actual image capturing (P = 0.063).
DISCUSSION This study evaluated 4 commercially available noncontact specular microscopes for quantitative and morphometric assessment of corneal endothelial cells in healthy and compromised corneas.
Successful Scans All 4 investigated instruments were highly successful in capturing and analyzing endothelial photographs in healthy
TABLE 1. Summary of Success Rates of Investigated Instruments (a Minimum of 1 Successful Reading of 3 Attempts Qualified as Success) Instrument
Healthy Eyes
CellChek XL auto CellChek XL semiauto Perseus auto Perseus manual EM-3000 auto CEM-530 auto
100% 100% 100% 100% 100% 100%
(22/22) (22/22) (22/22) (22/22) (22/22) (22/22)
FED 18.8% 18.8% 0% 0% 6.3% 0%
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
(3/16) (3/16) (0/16) (0/16) (1/16) (0/16)
Corneal Grafts 77.4% 77.4% 87.1% 93.5% 80.6% 64.5%
(24/31) (24/31) (27/31) (29/31) (25/31) (20/31)
Total 71.0% 71.0% 71.0% 73.9% 69.6% 60.9%
(49/69) (49/69) (49/69) (51/69) (48/69) (42/69)
www.corneajrnl.com |
383
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Cornea Volume 34, Number 4, April 2015
Luft et al
TABLE 2. Quantitative and Morphological Endothelial Cell Parameters of the 39 Eyes With a Minimum of 1 Image Suitable for Automated Analysis Obtained With All 4 Instruments (22 Healthy Eyes and 17 DSAEK Eyes) Instrument CellChek XL auto CellChek XL semi-auto (gold standard) Perseus auto EM-3000 auto CEM-530 auto
ECD, Cells/mm2 2480 2089 2096 2063 2093
6 6 6 6 6
429* 742 654 631 551
CV 48 30 30 39 30
6 6 6 6 6
36* 5 4 7* 7
Hexagonality, % 41 49 55 45 67
6 6 6 6 6
9* 11 7* 7 8*
CCT, mm 568 568 556 548 558
6 6 6 6 6
48 48 51* 46* 49
NUM, Cells
REP, Cells/mm2
6 6 6 6 6
81.8 81.2 38.1 84.4 68.6
121 93 179 179 91
386 56 74 83 62
Instrument repeatability (REP; mean SD of 3 consecutive measurements) was assessed in the sample of 24 eyes with 3 consecutive scans suitable for automated analysis obtained with all 4 instruments (20 healthy eyes and 4 DSAEK eyes). *Statistically significant difference compared with the gold standard (P , 0.05). CCT, central corneal thickness; CV, coefficient of variation of cell area; NUM, number of cells analyzed per scan.
corneas. In contrast, we found particularly poor performance of all devices in eyes with FED. As suggested by Hara et al,8 confocal microscopy, which is a contact technique, may be superior to noncontact specular microscopy in imaging eyes affected by FED. In our hands, all 4 instruments performed fairly well in the cohort of eyes with corneal transplants. In a sample of 62 eyes with DSEK grafts, Price et al9 reported slightly higher success rates for the EM-3000 and
for a previously marketed predecessor device to the CellChek XL.
Comparative Analysis The predecessor instrument to the CellChek XL is a popular choice for FDA clinical trials.10,11 Hence, for comparative analysis, we defined the semiautomated
FIGURE 1. Bland–Altman plots showing the differences in ECD measurements of the 4 instruments compared with the gold standard plotted against the mean value of corresponding measurements. The solid line represents the mean difference, and the dotted lines the 95% limits of agreement. A, CellChek XL with automated analysis minus gold standard. B, Perseus minus gold standard. C, EM-3000 minus gold standard. D, CEM-530 minus gold standard.
384
| www.corneajrnl.com
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Cornea Volume 34, Number 4, April 2015
Comparison of 4 Specular Microscopes
standard. However, in consideration of the commonly high endothelial cell levels in healthy eyes, these differences were of subclinical significance with mean differences of less than 200 cells per square millimeter and very few outliers, as shown in Figures 1B–D. Despite the statistically insignificant difference between automated and semiautomated analysis with the CellChek XL, these results must be interpreted with care. As apparent from the Bland–Altmann plot of agreement shown in Figure 1A, semiautomated analysis was performed in only 3 eyes in this subgroup with the CellChek XL. In all 3 cases, automated analysis considerably overstated ECD, in 1 case by more than 1700 cells per square millimeter. Overestimating ECD may adversely affect treatment decisions and could give rise to intraocular surgery-related complications as for instance corneal decompensation. Again, this obliges the clinician to meticulously assess the results obtained from automated image analysis programs.
DSAEK Eyes FIGURE 2. Box plots comparing the mean times taken for data entry, scanning, and automated image analysis with the 4 instruments.
endothelial cell analysis with the CellChek XL as the gold standard instrument. In the entire sample of eyes eligible for comparative analysis, no significant differences in the endothelial cell count were observed between the 4 devices. However, automated image analysis with the CellChek XL overstated ECD compared with the gold standard. This was predominantly due to considerable overestimation of ECD in the subgroup of DSAEK eyes. However, also in eyes without any detected corneal pathologies, the ECD was overestimated in 3 cases. We found that the software will often produce ECD readings within the reference range even in corneas with severe polymegethism, in corneas with very large endothelial cells or with very few or without any visible well-defined cells at all. Therefore, when using the CellChek XL automated image analysis program, it is essential to assess whether the software has correctly identified visible endothelial cells. There was little consistency between the 4 devices with respect to qualitative endothelial cell parameters and corneal pachymetry. Only the Perseus and the CEM-530 produced CV readings that were comparable with the gold standard. The EM-3000 was the only instrument that provided hexagonality readings interchangeable with those obtained with the gold standard device. Only small and clinically insignificant differences in pachymetry readings were observed between the investigated instruments with mean differences of 20 mm and less.
Healthy Corneas All instruments with the exception of the CellChek XL (automated analysis) significantly understated ECD in the subgroup of healthy corneas compared with the gold Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
In the DSAEK subgroup, automated analysis programs of the CellChek XL, the Perseus, and the CEM-530 overstated ECD. Previous studies have found that automated analysis programs are not sufficient in providing ECD measurements that are interchangeable with semiautomated or manual cellcounting methods.12–15 However, we found that the EM-3000 automated software provided quantitative endothelial measurements that were well comparable with those obtained with the manual gold standard method in the subgroup of post-DSAEK corneas. Identical findings were made in a study by Price et al,9 demonstrating interchangeability between manual methods of quantitative endothelial cell analysis and the EM-3000 automated software in post-DSEK eyes.
Repeatability All investigated specular microscopes provided quantitative endothelial cell measurements that were satisfactorily repeatable, and no statistically significant differences in repeatability were observed. This is relevant for follow-up measurements especially for phakic implants in patients with a healthy endothelium. In an earlier study, Salvetat et al16 found low intraobserver and interobserver test–retest variability for automated image analysis with the EM-3000 in a sample of 42 healthy eyes.
Time Measurements The fastest measurements were with the CEM-530 and the Perseus. The EM-3000 and the CellChek XL were slower. The time of the actual image capturing process was similar with all devices; hence, the differences in the total measurement time resulted from variations in the times taken for patient data entry and automated image analysis. Data entry with the EM-3000 and automated image analysis with the CellChek XL took longer than with the other devices as more clicks were necessary for these processes. Time taken for assessment is relevant when done routinely in patients planned to undergo either cataract of refractive surgery. www.corneajrnl.com |
385
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Cornea Volume 34, Number 4, April 2015
Luft et al
In conclusion, the CEM-530 and Perseus were extremely fast, showed highly repeatable measurements, and are recommendable for screening purposes such as preoperative assessment of the corneal status in a high-volume cataract clinic. The CellChek XL device took longer, but extensive postprocessing and an exportable database make this instrument particularly suitable for surveillance of compromised or transplanted corneas and for research purposes. The EM-3000 device was found to range between the fast devices and the CellChek XL and comes with reliable automated cell-counting software. Nevertheless, care should be taken when automated image analysis programs are applied. REFERENCES 1. Corneal endothelial photography. Three-year revision. American Academy of Ophthalmology. Ophthalmology. 1997;104:1360–1365. 2. Alio JL, Abbouda A, Pena-Garcia P, et al. Follow-up study of more than 15 years of an angle-supported phakic intraocular lens model (ZB5M) for high myopia: outcomes and complications. JAMA Ophthalmol. 2013; 131:1541–1546. 3. Neumayer T, Prinz A, Findl O. Effect of a new cohesive ophthalmic viscosurgical device on corneal protection and intraocular pressure in smallincision cataract surgery. J Cataract Refract Surg. 2008;34:1362–1366. 4. Takacs AI, Kovacs I, Mihaltz K, et al. Central corneal volume and endothelial cell count following femtosecond laser-assisted refractive cataract surgery compared to conventional phacoemulsification. J Refract Surg. 2012;28:387–391. 5. Villalba R, Jimenez A, Fornes G, et al. Flex center method versus center method for endothelial corneal evaluation in eye banking. A comparative analysis. Cell Tissue Bank. 2014;15:507–512.
386
| www.corneajrnl.com
6. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1: 307–310. 7. Adamis AP, Filatov V, Tripathi BJ, et al. Fuchs’ endothelial dystrophy of the cornea. Surv Ophthalmol. 1993;38:149–168. 8. Hara M, Morishige N, Chikama T, et al. Comparison of confocal biomicroscopy and noncontact specular microscopy for evaluation of the corneal endothelium. Cornea. 2003;22:512–515. 9. Price MO, Fairchild KM, Price FW Jr. Comparison of manual and automated endothelial cell density analysis in normal eyes and DSEK eyes. Cornea. 2013;32:567–573. 10. Jabbur NS. Endothelial cell studies in patients after photorefractive keratectomy for hyperopia. J Refract Surg. 2003;19:142–148. 11. McCarey BE, Edelhauser HF, Lynn MJ. Review of corneal endothelial specular microscopy for FDA clinical trials of refractive procedures, surgical devices, and new intraocular drugs and solutions. Cornea. 2008; 27:1–16. 12. Goldich Y, Marcovich AL, Barkana Y, et al. Comparison of corneal endothelial cell density estimated with 2 noncontact specular microscopes. Eur J Ophthalmol. 2010;20:825–830. 13. Imre L, Nagymihaly A. Reliability and reproducibility of corneal endothelial image analysis by in vivo confocal microscopy. Graefes Arch Clin Exp Ophthalmol. 2001;239:356–360. 14. Kitzmann AS, Winter EJ, Nau CB, et al. Comparison of corneal endothelial cell images from a noncontact specular microscope and a scanning confocal microscope. Cornea. 2005;24:980–984. 15. Klais CM, Buhren J, Kohnen T. Comparison of endothelial cell count using confocal and contact specular microscopy. Ophthalmologica. 2003;217:99–103. 16. Salvetat ML, Zeppieri M, Miani F, et al. Comparison between laser scanning in vivo confocal microscopy and noncontact specular microscopy in assessing corneal endothelial cell density and central corneal thickness. Cornea. 2011;30:754–759.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Comparison of 4 Specular Microscopes in Healthy Eyes and Eyes With Cornea Guttata or Corneal Grafts Nikolaus Luft, MD,* Nino Hirnschall, MD, PhD,* Sandra Schuschitz, MSc,* Petra Draschl, MD,* and Oliver Findl, MD, MBA*†
Purpose: The aim of this study was to compare 4 noncontact specular microscopes for the assessment of the corneal endothelium in a heterogeneous sample of eyes.
Methods: This prospective study included 22 healthy eyes, 16 eyes with cornea guttata and 31 eyes that had undergone penetrating keratoplasty or Descemet stripping automated endothelial keratoplasty. Corneal endothelial cell parameters of all eyes were assessed with the CellChek XL (Konan Medical, Hyogo, Japan), Perseus (Bon Optic, Lübeck, Germany), EM-3000 (Tomey, Nagoya, Japan), and CEM-530 (Nidek Co, Ltd, Gamagori, Japan) in a randomized order. Bland–Altman plots of interdevice agreement were produced. The time taken for data entry, scanning, and automated image analysis was recorded, and instrument repeatability was analyzed. Results: The mean age was 61 6 19 years (range, 24–88). The measurement success rate was 100% in healthy corneas, ranged between 64.5% and 93.5% in eyes with corneal grafts, and between 0% and 18.8% in eyes with cornea guttata. Comparative analysis of endothelial cell parameters revealed significant differences in endothelial cell density readings in the total study population and in subgroups (P , 0.01). Conclusions: The CEM-530 and Perseus were extremely fast, showed highly repeatable measurements, and are recommendable for screening purposes. The CellChek XL device was slower, but extensive postprocessing options and an exportable database make this instrument suitable for compromised corneas and for research purposes. The EM-3000 device ranged between the fast devices and the CellChek XL and comes with reliable automated cell-counting software. However, careful interpretation of results from automated image analysis software is mandatory. Key Words: specular microscopy, endothelial cell density, dystrophy, guttata (Cornea 2015;34:381–386)
Received for publication August 15, 2014; revision received December 28, 2014; accepted December 29, 2014. Published online ahead of print February 19, 2015. From the *VIROS—Vienna Institute for Research in Ocular Surgery, A Karl Landsteiner Institute, Hanusch Hospital, Vienna, Austria; and †Moorfields Eye Hospital NHS Foundation Trust, London, United Kingdom. The authors have no funding or conflicts of interest to disclose. Reprints: Oliver Findl, MD, MBA, Hanusch Hospital, Heinrich-CollinStrasse 30, 1140 Vienna, Austria (e-mail: oliver@findl.at). Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Cornea Volume 34, Number 4, April 2015
B
ecause of the inability of corneal endothelial cells to regenerate, it is paramount to detect patients with endothelial cell dystrophies or low levels of endothelial cells in routine preoperative testing, such as before cataract surgery.1 Moreover, reliable assessment of corneal endothelial health is mandatory in surveillance of patients with phakic intraocular lenses.2 A further useful application for corneal endothelial cell measurements is studies evaluating the effect of surgical techniques or devices on the corneal endothelium.3,4 Specular microscopes allow a quantitative and morphological assessment of corneal endothelial cells. The purpose of this study was to compare 4 commercially available, noncontact specular microscopes in a heterogeneous sample of eyes.
MATERIALS AND METHODS This prospective study included patients who had undergone Descemet stripping automated endothelial keratoplasty (DSAEK) or penetrating keratoplasty (PKP), patients with cornea guttata or Fuchs endothelial dystrophy (FED), and subjects with healthy corneas. Eyes with severe corneal scarring were excluded from this study. All the research and measurements followed the tenets of the Declaration of Helsinki, and all participants gave informed consent before the start of the study. Three examiners who were inexperienced in using all 4 investigated noncontact specular microscope systems, namely CellChek XL (Konan Medical, Hyogo, Japan), Perseus (Bon Optic, Lübeck, Germany), EM-3000 (Tomey, Nagoya, Japan), and CEM-530 (Nidek Co, Ltd, Gamagori, Japan), received an introduction for each of the 4 devices by experienced examiners. On the next day, both eyes of each subject were measured with the 4 specular microscopes in a single office visit. The instruments were used in a randomized order and 3 consecutive measurements of each eye were performed. Each patient was randomly assigned to 1 of the 3 examiners who then performed all measurements with all devices on that particular patient. After each scan, the patient was asked to lean back and the chinrest was readjusted. In all cases, the patient fixated on a central fixation light. In case of an acquired endothelial image without visible, well-defined cells, the scan was classified as unsuccessful. The following parameters were recorded for each eye: endothelial cell density (ECD) (cells/mm2), coefficient of variation of the cell area (CV), cell hexagonality (proportion of hexagonal cells in percentage), central corneal thickness www.corneajrnl.com |
381
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Luft et al
(mm), and the number of cells analyzed. Furthermore, the time taken for data entry, scanning, and automated image analysis was recorded for one randomly elected eye of each patient using a digital stopwatch. The timed examination was the first of 3 examinations on the respective eye in all cases. Additionally, a comprehensive anterior segment slit-lamp examination of all subjects was conducted at the beginning of their visit.
CellChek XL The CellChek XL specular microscope system captures a photographic field of 0.1 mm2. After subsequent autoalignment and automated image acquisition, image analysis was performed using the built-in automated cell-counting software in default settings. For autoanalysis, the software uses predefined reference patterns with different cell sizes (“Small,” “Medium,” “Large,” and “XLarge”) and defaults to “Small.” The autoanalyzed image was saved electronically to a database implemented in the CellChek XL system. Additionally, manual (“semiautomated”) evaluation of all endothelium photographs was performed. Different editing tools of the onboard software were used as advised in the instruction manual: selecting an appropriate cell size reference pattern, resizing the area to be analyzed by the autodetection algorithm, and the “Center method” or “Flex-Center method”5 for manual cell counting.
Perseus With the Perseus specular microscope, up to 300 endothelial cells can be counted within an area of 0.2 mm2. The Perseus provides a touch screen to roughly align the eye to be examined. After subsequent automated alignment and autoacquisition, the captured image was analyzed by the cellidentification software and saved to the onboard database. In case of images with visible cells that were missed by the cellidentification algorithm, the manual analysis method for identifying endothelial was performed instead in accordance with the instruction manual: cell corner tracing with a touch pen, erasing incorrectly traced cell borders, excluding areas from analysis that are obscured (eg, by guttae).
EM-3000 The EM-3000 specular microscope uses an optical magnification of ·190 and up to 300 cells per image are counted within an area of 0.1 mm2. This device automatically captured 15 images per measurement, which were displayed and sorted by the software with respect to image quality. The image with highest quality was then selected by the examiner and automated cell detection, and counting was performed using the built-in manufacturer’s software. Manual image analysis was not performed with the EM-3000 because the study device was not supplied with the proprietary database that allows manual image after processing.
Cornea Volume 34, Number 4, April 2015
captures a total of 16 endothelial photographs per scan, which are displayed and automatically sorted based on quality immediately after scanning. The most suitable image according to the judgment of the examiner was elected for automated cell detection with the manufacturer’s software. Because the CEM-530 offered only a very basic method for manual cell detection, we abstained from performing manual analysis with this machine.
Statistical Analysis All statistical analysis was performed using Excel Version 14.4.1 for Mac (Microsoft Corporation) and SPSS Statistics Version 21.0.0.0 for Mac (IBM). The Kolmogorov–Smirnov test was used to confirm the normality of data distribution. Descriptive data are always presented as mean, SD, and range. P , 0.05 was considered statistically significant. A Fisher exact test with a Bonferroni correction was used for multiple comparisons of measurement success rate between devices (nonmetric data). For comparative statistical analysis of endothelial cell parameters, the median value of 3 consecutive measurements of 1 device was taken into account. In the case of 2 successful scans, the mean value was used instead and if only 1 measurement was successful, this measurement was used for further analysis. A repeatedmeasures analysis of variance (ANOVA) with a Bonferroni correction was used to compare ECD, CV, cell hexagonality, and central corneal thickness readings between instruments and to determine whether these differed significantly. Semiautomated analysis with the CellChek XL was defined as the gold standard for quantitative and qualitative assessment of the endothelial cells. Agreement of ECD readings between the respective devices was evaluated using the Bland–Altman method.6 The limits of agreement (LoA) were defined as the mean interdevice difference 6 2 SD. For repeatability testing, a 1-way ANOVA was applied to compare mean SD of 3 consecutive ECD readings between devices. A Friedman test and post hoc analysis with Wilcoxon signed-rank test conducted with a Bonferroni correction were performed to assess differences in time measurements between the instruments.
RESULTS
The CEM-530 noncontact specular microscope captures an endothelial image 0.1 mm2 in size. This device
The study included 69 eyes of 36 patients: 22 healthy eyes, 16 eyes with FED, and 31 eyes with corneal grafts (29 DSAEK eyes and 2 PKP eyes). Male to female ratio was 15 (42%) to 21 (58%). The mean age was 61 6 19 years (range, 24–88). Three eyes were excluded from the study: the first was an eye that had undergone PKP with a decompensated cornea and phthisis. The second had end-stage corneal dystrophy, severe stromal vascularization, and calcification. In the third case, the subject withdrew from the trial for personal reasons before measurements with all 4 devices could be completed. For semiautomated analysis of the acquired images with the CellChek XL, cell size patterns other than the default “small” pattern were selected in 36 scans of 16 eyes. In 29 scans of 15 eyes, the area of analysis in the endothelial
382
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
CEM-530
| www.corneajrnl.com
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Cornea Volume 34, Number 4, April 2015
Comparison of 4 Specular Microscopes
photograph was cropped. Moreover, the “Center method” was applied in 29 scans of 12 eyes, and the “Flex-center” method was used in 12 scans of 6 eyes. With these means of manual image analysis, no additional scans were obtainable compared with automated analysis. With the Perseus, manual image analysis was performed on 5 endothelial photographs of 2 eyes by tracing the cell borders. In neither of these 2 eyes, successful scans had been obtainable with automated image analysis. A summary of successful endothelial cell measurements with the 4 instruments is given in Table 1. No significant difference in the success rate of measurements between instruments was observed (all with P . 0.05).
Comparative Analysis
and the CEM-530 were significantly higher compared with the gold standard with mean differences of 681, 166, and 262 cells per square millimeter, respectively (all with P , 0.01). In DSAEK eyes, only the EM-3000 produced ECD readings that were comparable with the gold standard device (mean difference, 119 cells/mm2; P = 0.31).
Repeatability The mean SD of 3 consecutive ECD measurements with each instrument are shown in Table 2. A 1-way ANOVA revealed no statistically significant differences in repeatability between the 4 instruments (P = 0.304).
Time Measurements
A minimum of 1 image suitable for automated analysis was obtainable with all 4 instruments in 39 eyes of 26 patients (22 healthy eyes and 17 DSAEK eyes). With the Perseus, manual image analysis was performed in none of these eyes, and thus, no differentiation between manual and automated analysis was made for this instrument in comparative analysis. The quantitative and morphological endothelial cell parameters measured with each device are summarized in Table 2. Mean ECD measurements obtained with the Perseus, the EM-3000, or the CEM-530 did not differ significantly compared with those obtained with the gold standard device with mean differences of 6, 26, and 4 cells per square millimeter, respectively (all with P = 0.99). In contrast, automated image analysis with the CellChek XL produced ECD readings that were significantly higher (mean difference, 391 cells/mm2; P , 0.01). Bland–Altman plots of interdevice differences (Fig. 1) also demonstrate that the difference in ECD readings was largest between automated image analysis with the CellChek XL and the gold standard. In the subgroup of healthy corneas, no significant differences in ECD readings were detected between the Perseus, the EM-3000, and the CEM-530. However, these 3 instruments understated ECD compared with the gold standard with mean differences of 117, 137, and 196 cells per square millimeter, respectively (all with P , 0.01). Automated analysis with the CellChek XL produced readings comparable with the reference device with a mean difference of 167 cells per square millimeter (P = 0.99). However, ECD readings with these 2 methods only differed in 3 cases (Fig. 1A). In the DSAEK cohort, ECD readings obtained with the CellChek XL using automated image analysis, the Perseus,
The mean total measurement time was 72 6 20 seconds (range, 46–121) with the CellChek XL, 54 6 21 seconds (range, 36–160) with the Perseus, 64 6 19 seconds (range, 36– 144) with the EM-3000, and 53 6 12 seconds (range, 37–91) with the CEM-530. The Friedman test and a post hoc Wilcoxon signed-rank test conducted with a Bonferroni correction applied (resulting in a significance level set at P , 0.0083) revealed no significant difference in mean total measurement time between the Perseus and the CEM-530 (P = 0.75). Equally, no difference was observed between the CellChek XL and the EM-3000 (P = 0.039). Measurements with the latter devices took significantly longer than with the former 2 (all with P , 0.003). Box plots illustrating the mean times taken for data entry, scanning, and automated image analysis with the 4 instruments are shown in Figure 2. Significant differences between devices were discovered in the mean times taken for patient data entry and for automated image analysis (P , 0.001). No significant difference was found in the mean times taken for actual image capturing (P = 0.063).
DISCUSSION This study evaluated 4 commercially available noncontact specular microscopes for quantitative and morphometric assessment of corneal endothelial cells in healthy and compromised corneas.
Successful Scans All 4 investigated instruments were highly successful in capturing and analyzing endothelial photographs in healthy
TABLE 1. Summary of Success Rates of Investigated Instruments (a Minimum of 1 Successful Reading of 3 Attempts Qualified as Success) Instrument
Healthy Eyes
CellChek XL auto CellChek XL semiauto Perseus auto Perseus manual EM-3000 auto CEM-530 auto
100% 100% 100% 100% 100% 100%
(22/22) (22/22) (22/22) (22/22) (22/22) (22/22)
FED 18.8% 18.8% 0% 0% 6.3% 0%
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
(3/16) (3/16) (0/16) (0/16) (1/16) (0/16)
Corneal Grafts 77.4% 77.4% 87.1% 93.5% 80.6% 64.5%
(24/31) (24/31) (27/31) (29/31) (25/31) (20/31)
Total 71.0% 71.0% 71.0% 73.9% 69.6% 60.9%
(49/69) (49/69) (49/69) (51/69) (48/69) (42/69)
www.corneajrnl.com |
383
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Cornea Volume 34, Number 4, April 2015
Luft et al
TABLE 2. Quantitative and Morphological Endothelial Cell Parameters of the 39 Eyes With a Minimum of 1 Image Suitable for Automated Analysis Obtained With All 4 Instruments (22 Healthy Eyes and 17 DSAEK Eyes) Instrument CellChek XL auto CellChek XL semi-auto (gold standard) Perseus auto EM-3000 auto CEM-530 auto
ECD, Cells/mm2 2480 2089 2096 2063 2093
6 6 6 6 6
429* 742 654 631 551
CV 48 30 30 39 30
6 6 6 6 6
36* 5 4 7* 7
Hexagonality, % 41 49 55 45 67
6 6 6 6 6
9* 11 7* 7 8*
CCT, mm 568 568 556 548 558
6 6 6 6 6
48 48 51* 46* 49
NUM, Cells
REP, Cells/mm2
6 6 6 6 6
81.8 81.2 38.1 84.4 68.6
121 93 179 179 91
386 56 74 83 62
Instrument repeatability (REP; mean SD of 3 consecutive measurements) was assessed in the sample of 24 eyes with 3 consecutive scans suitable for automated analysis obtained with all 4 instruments (20 healthy eyes and 4 DSAEK eyes). *Statistically significant difference compared with the gold standard (P , 0.05). CCT, central corneal thickness; CV, coefficient of variation of cell area; NUM, number of cells analyzed per scan.
corneas. In contrast, we found particularly poor performance of all devices in eyes with FED. As suggested by Hara et al,8 confocal microscopy, which is a contact technique, may be superior to noncontact specular microscopy in imaging eyes affected by FED. In our hands, all 4 instruments performed fairly well in the cohort of eyes with corneal transplants. In a sample of 62 eyes with DSEK grafts, Price et al9 reported slightly higher success rates for the EM-3000 and
for a previously marketed predecessor device to the CellChek XL.
Comparative Analysis The predecessor instrument to the CellChek XL is a popular choice for FDA clinical trials.10,11 Hence, for comparative analysis, we defined the semiautomated
FIGURE 1. Bland–Altman plots showing the differences in ECD measurements of the 4 instruments compared with the gold standard plotted against the mean value of corresponding measurements. The solid line represents the mean difference, and the dotted lines the 95% limits of agreement. A, CellChek XL with automated analysis minus gold standard. B, Perseus minus gold standard. C, EM-3000 minus gold standard. D, CEM-530 minus gold standard.
384
| www.corneajrnl.com
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Cornea Volume 34, Number 4, April 2015
Comparison of 4 Specular Microscopes
standard. However, in consideration of the commonly high endothelial cell levels in healthy eyes, these differences were of subclinical significance with mean differences of less than 200 cells per square millimeter and very few outliers, as shown in Figures 1B–D. Despite the statistically insignificant difference between automated and semiautomated analysis with the CellChek XL, these results must be interpreted with care. As apparent from the Bland–Altmann plot of agreement shown in Figure 1A, semiautomated analysis was performed in only 3 eyes in this subgroup with the CellChek XL. In all 3 cases, automated analysis considerably overstated ECD, in 1 case by more than 1700 cells per square millimeter. Overestimating ECD may adversely affect treatment decisions and could give rise to intraocular surgery-related complications as for instance corneal decompensation. Again, this obliges the clinician to meticulously assess the results obtained from automated image analysis programs.
DSAEK Eyes FIGURE 2. Box plots comparing the mean times taken for data entry, scanning, and automated image analysis with the 4 instruments.
endothelial cell analysis with the CellChek XL as the gold standard instrument. In the entire sample of eyes eligible for comparative analysis, no significant differences in the endothelial cell count were observed between the 4 devices. However, automated image analysis with the CellChek XL overstated ECD compared with the gold standard. This was predominantly due to considerable overestimation of ECD in the subgroup of DSAEK eyes. However, also in eyes without any detected corneal pathologies, the ECD was overestimated in 3 cases. We found that the software will often produce ECD readings within the reference range even in corneas with severe polymegethism, in corneas with very large endothelial cells or with very few or without any visible well-defined cells at all. Therefore, when using the CellChek XL automated image analysis program, it is essential to assess whether the software has correctly identified visible endothelial cells. There was little consistency between the 4 devices with respect to qualitative endothelial cell parameters and corneal pachymetry. Only the Perseus and the CEM-530 produced CV readings that were comparable with the gold standard. The EM-3000 was the only instrument that provided hexagonality readings interchangeable with those obtained with the gold standard device. Only small and clinically insignificant differences in pachymetry readings were observed between the investigated instruments with mean differences of 20 mm and less.
Healthy Corneas All instruments with the exception of the CellChek XL (automated analysis) significantly understated ECD in the subgroup of healthy corneas compared with the gold Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
In the DSAEK subgroup, automated analysis programs of the CellChek XL, the Perseus, and the CEM-530 overstated ECD. Previous studies have found that automated analysis programs are not sufficient in providing ECD measurements that are interchangeable with semiautomated or manual cellcounting methods.12–15 However, we found that the EM-3000 automated software provided quantitative endothelial measurements that were well comparable with those obtained with the manual gold standard method in the subgroup of post-DSAEK corneas. Identical findings were made in a study by Price et al,9 demonstrating interchangeability between manual methods of quantitative endothelial cell analysis and the EM-3000 automated software in post-DSEK eyes.
Repeatability All investigated specular microscopes provided quantitative endothelial cell measurements that were satisfactorily repeatable, and no statistically significant differences in repeatability were observed. This is relevant for follow-up measurements especially for phakic implants in patients with a healthy endothelium. In an earlier study, Salvetat et al16 found low intraobserver and interobserver test–retest variability for automated image analysis with the EM-3000 in a sample of 42 healthy eyes.
Time Measurements The fastest measurements were with the CEM-530 and the Perseus. The EM-3000 and the CellChek XL were slower. The time of the actual image capturing process was similar with all devices; hence, the differences in the total measurement time resulted from variations in the times taken for patient data entry and automated image analysis. Data entry with the EM-3000 and automated image analysis with the CellChek XL took longer than with the other devices as more clicks were necessary for these processes. Time taken for assessment is relevant when done routinely in patients planned to undergo either cataract of refractive surgery. www.corneajrnl.com |
385
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Cornea Volume 34, Number 4, April 2015
Luft et al
In conclusion, the CEM-530 and Perseus were extremely fast, showed highly repeatable measurements, and are recommendable for screening purposes such as preoperative assessment of the corneal status in a high-volume cataract clinic. The CellChek XL device took longer, but extensive postprocessing and an exportable database make this instrument particularly suitable for surveillance of compromised or transplanted corneas and for research purposes. The EM-3000 device was found to range between the fast devices and the CellChek XL and comes with reliable automated cell-counting software. Nevertheless, care should be taken when automated image analysis programs are applied. REFERENCES 1. Corneal endothelial photography. Three-year revision. American Academy of Ophthalmology. Ophthalmology. 1997;104:1360–1365. 2. Alio JL, Abbouda A, Pena-Garcia P, et al. Follow-up study of more than 15 years of an angle-supported phakic intraocular lens model (ZB5M) for high myopia: outcomes and complications. JAMA Ophthalmol. 2013; 131:1541–1546. 3. Neumayer T, Prinz A, Findl O. Effect of a new cohesive ophthalmic viscosurgical device on corneal protection and intraocular pressure in smallincision cataract surgery. J Cataract Refract Surg. 2008;34:1362–1366. 4. Takacs AI, Kovacs I, Mihaltz K, et al. Central corneal volume and endothelial cell count following femtosecond laser-assisted refractive cataract surgery compared to conventional phacoemulsification. J Refract Surg. 2012;28:387–391. 5. Villalba R, Jimenez A, Fornes G, et al. Flex center method versus center method for endothelial corneal evaluation in eye banking. A comparative analysis. Cell Tissue Bank. 2014;15:507–512.
386
| www.corneajrnl.com
6. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1: 307–310. 7. Adamis AP, Filatov V, Tripathi BJ, et al. Fuchs’ endothelial dystrophy of the cornea. Surv Ophthalmol. 1993;38:149–168. 8. Hara M, Morishige N, Chikama T, et al. Comparison of confocal biomicroscopy and noncontact specular microscopy for evaluation of the corneal endothelium. Cornea. 2003;22:512–515. 9. Price MO, Fairchild KM, Price FW Jr. Comparison of manual and automated endothelial cell density analysis in normal eyes and DSEK eyes. Cornea. 2013;32:567–573. 10. Jabbur NS. Endothelial cell studies in patients after photorefractive keratectomy for hyperopia. J Refract Surg. 2003;19:142–148. 11. McCarey BE, Edelhauser HF, Lynn MJ. Review of corneal endothelial specular microscopy for FDA clinical trials of refractive procedures, surgical devices, and new intraocular drugs and solutions. Cornea. 2008; 27:1–16. 12. Goldich Y, Marcovich AL, Barkana Y, et al. Comparison of corneal endothelial cell density estimated with 2 noncontact specular microscopes. Eur J Ophthalmol. 2010;20:825–830. 13. Imre L, Nagymihaly A. Reliability and reproducibility of corneal endothelial image analysis by in vivo confocal microscopy. Graefes Arch Clin Exp Ophthalmol. 2001;239:356–360. 14. Kitzmann AS, Winter EJ, Nau CB, et al. Comparison of corneal endothelial cell images from a noncontact specular microscope and a scanning confocal microscope. Cornea. 2005;24:980–984. 15. Klais CM, Buhren J, Kohnen T. Comparison of endothelial cell count using confocal and contact specular microscopy. Ophthalmologica. 2003;217:99–103. 16. Salvetat ML, Zeppieri M, Miani F, et al. Comparison between laser scanning in vivo confocal microscopy and noncontact specular microscopy in assessing corneal endothelial cell density and central corneal thickness. Cornea. 2011;30:754–759.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
