Endothelial Morphometry by Image Analysis of Corneas Organ Cultured at 31°C

Abstract

Purpose.: To determine the factors influencing endothelial morphometry by using image analysis of corneas stored in organ culture to determine the coefficient of variation (CV) in cell area and percentage of hexagonal cells.

Methods.: The endothelia of 505 of the 559 corneas consecutively stored at the eye bank were routinely analyzed with Sambacornea image-analysis software (ver. 1.2.10; Tribvn, Châtillon, France) on three large-field images of 750 × 1000 μm, obtained after osmotic dilation of the intercellular spaces with 0.9% sodium chloride. Analysis was performed on at least 300 cells. The quality of the three-image set was graded poor, average, or good by an independent observer. The studied parameters were donor age and sex, lens status, storage time, and intrinsic quality of captured images. Statistics were analyzed by nonparametric tests.

Results.: Image analysis was possible for 504 of the 505 assessed corneas. Donor age correlated significantly with endothelial cell density (ECD; r = −0.343), CV (r = 0.221), and hexagonality (r = −0.314; P < 0.001 for the three). Image quality significantly influenced these three parameters. ECD and hexagonality decreased parallel to image quality, whereas the CV increased. In the 258 corneas assessed twice (on average, at day [D] +4, then D +14) ECD, CV, and hexagonality decreased during storage.

Conclusions.: Despite the sometimes mediocre quality of the transmitted light microscopy images, endothelial parameters supplied by the analyzer were clinically reliable, since variations similar to those long known in specular microscopy were found. Endothelial morphometry (CV and hexagonality) is likely to provide further information on the endothelial function of the graft tissue, perhaps particularly for grafts of borderline ECD, close to the discard threshold.

Assessment of the endothelial quality of corneas stored in eye banks is the most important factor in the graft selection process, after microbiologic safety criteria. The human corneal endothelial mosaic is typically characterized by its endothelial cell density (ECD), which, if high, provides a significant functional reserve, and by cell morphology. Endothelial cell morphology is assessed by size uniformity (polymegethism) and cell shape (pleomorphism). The ideal endothelium has high ECD and low polymegethism and pleomorphism and therefore presents as a mosaic principally comprising small hexagons of identical size. In eye banks, these parameters are assessed by observing the endothelium with a microscope: either a specular microscope (SM) at the start of storage at 4°C, when the cornea is still sufficiently transparent and fine, or a bright-field or phase-contrast light microscope at any point during organ culture (OC). OC, the standard storage method in Europe, ¹ causes reversible stromal edema and a transient decrease in transparency, which preclude the use of specular microscopy. 

In eye bank history, many methods have successively been used to assess ECD and morphometry. Manual methods have always allowed ECD to be estimated quantitatively and expressed in number of cells per square millimeter. However, their reliability of these methods has been questioned in some cases. ^2,3 The development of computerized image analysis has made counting much more reliable. ^4–8 ECD is a quantitative criterion that has long been universally recognized as a condition of graft survival. Further, it has recently been shown, in a prospective study of a cohort of 500 graft recipients, that high ECD in the graft is the parameter that most directly retards the onset of graft opacification by late endothelial failure. ^9,10  

The utility of computerized image analysis for the objective quantification of endothelial morphometry was perceived very early on, after the development of SM, ^11,12 and then it became commonplace with the generalization of computers. Before SM was used, naked-eye observation of the endothelium allowed assessment of only polymegethism and pleomorphism, which could at best be graded in the categories low, moderate, and high. Image-analysis programs quantify morphometry of objects using many parameters, of which the most common are area, perimeter, length of sides (by calculating, for each, their mean and variance), number of neighbors, form factor, cell elongation, and deviation. ^11,13 In common practice, derived from clinical use of commercial SM and the accompanying analysis software programs, only two criteria are used: coefficient of variation (CV) of percentage of cell area = standard deviation of area/mean area (in square millimeters), and the percentage of hexagonal cells, often called hexagonality, even though these are probably too restrictive to exhaustively describe endothelial morphometry. ^13,14 The importance of the relationship between morphometry and endothelial cell function has been discussed for more than 25 years ^11,15,16 and direct involvement of morphometric modifications in edematous decompensation of the cornea has been demonstrated in pseudophakic bullous keratopathy ¹⁷ and primitive endothelial dystrophies. ^18,19 Conversely, it has been shown that large CV variations in long-term contact lens wearers have no impact on endothelial permeability and its deswelling capacity. ²⁰ The influence of morphometric parameters on graft endothelium quality is unknown. To date, it has only been recommended that graft tissues presenting high polymegethism be excluded. ²¹ We thought it was of paramount importance to determine whether there are pleomorphism and polymegethism thresholds that would influence graft survival that could constitute meeting graft discard criteria. We believe that this question is particularly crucial for graft tissue with ECD close to the discard threshold (generally, 2000–2400 cells/mm², depending on the bank). To address this point, we thought it was first necessary to know the criteria variation spectrum given by endothelium assessment tools in eye banks and to ensure that these criteria were valid. Unlike storage at 4°C, for which specular image analyzers have long been used and the morphometric criteria are known, the routine image analysis of optic microscope micrographs obtained in OC is nascent in Europe. Last, even in the case of SM, these thresholds are not yet defined. Even in large-scale American clinical trials such as the Cornea Donor Study (CDS), graft inclusion criteria relative to endothelial morphometry are still subjective: Pleomorphism and polymegethism must be no more than moderate. 

The purpose of this work was twofold: to analyze the relevance of endothelial morphometry data calculated by image analysis during OC and to identify the parameters that influence them. 

Material and Methods

Retrieval and OC

During 20 months, 559 corneas of 280 donors (one donor with a sole eye) were received at the eye bank of the French Blood Center (Loire/Auvergne region). Retrieval was performed by in situ excision of the corneoscleral rim in the morgues of our university hospital and of two general hospitals by 12 ophthalmologists, 8 of whom were ophthalmology department residents. No donor age limit was set. The retrievers observed each cornea macroscopically to grade its quality as very good, good, or average, depending on epithelial abrasions, stromal aspect (diffuse edema and localized opacities), endothelial folds, and presence of an arcus senilis. The presence of an intraocular lens (IOL) was noted. Corneas were immediately placed in OC medium (CorneaMax; Eurobio, Les Ulis, France) in a temperature-controlled incubator at 31°C. The OC modus operandi was as follows: initial endothelial assessment in the first days of storage with renewal of medium; further renewal 14 days later, as necessary; graft scheduling; final endothelial assessment 2 days before grafting; and immersion in deswelling medium. Corneas with initial ECD lower than 2000 cells/mm² were not reassessed. 

Endothelial Assessments

The two endothelial analyses were conducted using the same protocol. The cornea extracted from its medium was briefly rinsed by gentle sprinkling on both sides with saline buffer (Balanced Salt Solution, BSS plus; Alcon, Rueil Malmaison, France), and the endothelial face was covered with 0.4% trypan blue (Eurobio) for 1 minute, then with 0.9% sodium chloride (Aguettant, Lyon, France) for 4 minutes, renewed every minute. The cornea was placed in a sealed Petri dish and observed by optical microscope (Laborlux S; Leica, Wetzlar, Germany) fitted with a digital camera (DXC 755P; Sony, Tokyo, Japan). A series of 10 nonoverlapping photographs was taken with a ×10 lens in the 8-mm diameter center of the cornea. Large-field images (750 × 1000 μm) were acquired directly by the Sambacornea software program (ver. 1.2.10; Tribvn, Châtillon, France) with 768 × 576-pixel resolution and 8-bit gray level in bitmap format. On the 20-in. screen routinely used, total magnification of displayed images was approximately ×220. The entire system was calibrated horizontally and vertically with a calibration micrometer (Leica Microsystems, Rueil-Malmaison, France) by entering the software's known dimensions and checked by measuring the ECDs of a specific calibration blade designed for eye banks and described elsewhere. ⁸ This two-dimensional calibration process was used to take account of any distortions in magnification that could lead to imperfectly square pixels, a problem that we had observed with several digital cameras. The actual analysis was conducted by using a procedure validated ²² by either of the eye bank's two full-time technicians, each of whom had more than 3 years' experience including more than 1000 endothelial cell counts. After choosing the three best available images, the technician delimited the areas of interest to be counted, avoiding areas of endothelial folding liable to promote parallax errors, and then corrected any boundaries incorrectly recognized by the software. Unless ECD was too low, 300 cells at most were counted. The calculated parameters were the number of cells considered for ECD calculation, ECD (cells/mm²), CV (%), number of cells considered for calculation of polygonality, and percentages of hexagons, pentagons, and heptagons. 

Image Quality

The quality of the three images in each analysis was graded by an observer (AD) unaware of the donor's characteristics and of the image-analysis result, using a three-level score previously described ²³ and based on visualization of cell borders, background noise, and surface area of clearly visible cells (Fig. 1). The score was 2 if the cell borders were easily visible, with little or no background noise, over at least two thirds of the image; 1 if the cell borders were easily visible, with moderate background noise, over one third to two thirds of the image; and 0 if the cell borders were hard to visualize, with high background noise, over less than one third of the image. Based on the three scores, overall quality of the three-image set was graded as poor (0/0/0, 0/0/1, or 0/1/1), average (1/1/1, 0/1/2, or 2/1/1), or good (2/2/1 or 2/2/2). Other combinations were not observed. 

Figure 1.

 

Endothelial image quality during OC. Representative example of sets of three images. Visualization of cell borders, background noise, and surface of area of clearly visible cells were taken into account. (A) Good, (B) average, and (C) poor quality. Scale bar, 200 μm.

Statistical Analysis

Unless otherwise stated, data are expressed as the mean (SD) or median (range), depending on their distribution. Data distribution normality was tested with the Kolmogorov-Smirnov test (Lilliefors variant) and the Shapiro-Wilk test, with a nonnormality threshold of P < 0.05. The mean values of the quantitative variables were compared by the paired t-test if distribution was normal and otherwise with the Wilcoxon or the Kruskal-Wallis nonparametric tests. Qualitative variables were compared by the χ² test. Relations between donor age and endothelial analysis parameters were sought by calculating the Pearson correlation coefficient r. To study relations between image-set quality and endothelial parameters, the results of the analyses at the start and end of OC were grouped. The significance threshold was set at P < 0.05 (Statistical Package for Social Sciences for Windows ver. 11.5.1; SSPS, Chicago, IL). 

Results

Donor Characteristics

The mean (SD) age of the 280 donors was 73 (15) years. All were Caucasian. The process used with the 559 corneas is presented in Figure 2. In total, 505 corneas from 253 donors, aged 72 (15) years, underwent at least one endothelial analysis by the analyzer. 

Figure 2.

 

Flow chart of analyses of 559 consecutive corneas received at the eye bank during the prospective study.

Corneas from heart-beating donors were retrieved at the end of organ retrieval (i.e., in the very first minutes after circulatory arrest). Postmortem time for the other donors was 15 (range, 3–31) hours. Note that 8.9% were retrieved within 6 hours. The other characteristics are presented in Table 1. 

Table 1.

 

Characteristics of the 253 Donors from Whom at Least One Cornea Underwent Endothelial Analysis

	n	%
Age classes, y
<55	34	13
55–79	124	49
≥80	95	38
Sex
Male	170	67
Type of death
Non-heart-beating	230	91
Cause of death
Cancer	84	33.2
Cardiac	74	29.2
Stroke	32	12.6
Respiratory	31	12.3
Infection	5	2.0
Trauma	2	0.8
Poisoning	1	0.4
Other	23	9.1
Unknown*	1	0.4
Lens status (n = 505)
Phakic	440	87
PCL phacoemulsification	43	9
PCL manual	5	1
ACL and aphakic	5	1
Not specified	11	2

 PCL, posterior chamber lens; ACL: anterior chamber lens.

*  Natural death at 88 years.

Endothelial Characteristics at the Start of Storage in OC

The first assessment was performed at 4 (range, 1–12) days of OC. Only one cornea (0.2%) could not be analyzed because of poor-quality images of an aphakic eye that was found several days later to be contaminated. The overall characteristics of the initial endothelial analysis of 504 corneas are detailed in Table 2. 

Table 2.

 

Endothelial Characteristics of 504 Corneas Analyzed at the Start of OC

	Mean	SD	Min	Max	Median
Cells, n	318	39	105	436	317
ECD, cells/mm2	2672	721	677	4281	2752
CV, %	29.4	6.3	16.4	60.6	28.4
Pentagons, %	25.5	4.9	10.7	58.8	25.4
Hexagons, %	52.4	8.9	17.6	80.6	52.5
Heptagons, %	17.0	3.6	5.8	28.7	17.15

Significant correlations were found between age, ECD, CV, and the three polygonality criteria. With age, there was a decrease in ECD (r = −0.345) and in hexagonality (r = −0.314), and, conversely, an increase in CV (r = 0.225) and the percentage of pentagons (r = 0.231) and heptagons (r = 0.214) (P < 0.01 for each of the five items). None of the parameters differed between sexes. A moderate but significant degree of correlation was also found between ECD and CV (r = −0.402) and hexagonality (r = 0.419; P < 0.01 for each; Fig. 3). 

Figure 3.

 

ECD, CV, and percentage of hexagonal cells according to donor age. CV and percentage of hexagonal cells according to ECD (P < 0.01 for all five; N = 504).

The 46 corneas retrieved from heart-beating donors had significantly better endothelial parameters (data not shown) than did other corneas, but the donors' ages were significantly lower: 48 (range, 34–77) years versus 78 (range, 16–100) years (P < 0.01). 

Endothelial parameters differed in the three classes, determined by the retriever's subjective judgment, with ECD and hexagonality increasing significantly and CV and percentage of pentagonal cells decreasing when the stromal aspect was better (data not shown). However, this judgment was also linked to donor age, with 78 (12) years for corneas graded average, 74 (14) years for corneas graded good, and 68 (16) years for corneas graded very good (P < 0.01). 

Given the low number of aphakic or anterior chamber lenses (n = 4), these subpopulations were not studied. Compared with postsurgical cataractous eyes with posterior chamber lenses, corneas from phakic eyes had higher ECD and hexagonality and lower CV and percentages of pentagons and heptagons (Table 3). The age of postsurgical donors was significantly higher than that of nonsurgical donors: respectively, 87 (56–98) and 74 (16–100) years (P < 0.01). 

Table 3.

 

Endothelial Characteristics at the Start of OC of Corneas from Phakic Eyes and from Postsurgical Cataractous Eyes with Posterior Chamber Lenses

	Mean	SD	Min	Max	Median	P
Cells, n
Phakic	320	35	125	417	318	0.062
PCL	312	46	136	436	312
ECD, cells/mm2
Phakic	2720	683	677	4281	2788	<0.01
PCL	2216	808	805	3852	2147
CV, %
Phakic	29.2	6.3	16.4	60.6	28.3	0.024
PCL	31.1	6.0	21.6	44.2	31.2
Pentagons, %
Phakic	25.4	4.8	10.7	41.1	25.3	0.176
PCL	26.8	6.6	17.0	58.8	27.0
Hexagons, %
Phakic	52.8	8.8	21.5	80.6	53.2	0.024
PCL	49.4	9.2	17.6	71.0	48.4
Heptagons, %
Phakic	17.0	3.5	7.8	28.7	17.0	0.063
PCL	17.6	3.4	7.0	23.3	18.1

 Phakic eyes (n = 440); posterior chamber lens (PCL; n = 48).

Evolution of Morphometry during OC

Two hundred fifty-eight corneas (51% of the 504, 55% of the phakic eyes, and only 27% of the eyes with IOLs) underwent a second analysis at the end of OC, a mean of 9 days (range, 2–24) after the first examination. The mean age of the 154 donors in this subgroup was 69 years (range, 16–99). The evolution of the endothelial analysis parameters is detailed in Table 4. ECD, CV, and hexagonality decreased significantly. Percentages of pentagons and heptagons increased. Daily cell loss was −1.28% (range, −10.4% to −2.8%). 

Table 4.

 

Evolution of Endothelial Analysis Characteristics during OC

	Mean	SD	Min	Max	Median	P
Cells, n
Start	329	30	154	417	322	<0.01
End	317	20	199	420	315
ECD, cells/mm2
Start	3099	422	2152	4281	3083	<0.01
End	2706	399	1451	3969	2679
CV, %
Start	27.1	4.3	16.4	40.7	26.7	<0.01
End	25.7	3.4	18.1	36.3	25.4
Pentagons, %
Start	24.1	4.3	10.7	39.5	24.1	<0.01
End	26.2	4.1	12.8	37.8	26.1
Hexagons, %
Start	55.7	7.7	38.7	80.6	56.0	<0.01
End	50.7	6.8	30.6	71.0	51.0
Heptagons, %
Start	16.5	3.3	7.0	26.6	16.7	<0.01
End	17.8	3.6	5.2	33.3	18.0

  n = 258.

Endothelial Characteristics of the Subgroup of Corneas Excluded for Endothelial Reasons after the First Analysis

Sixty-eight corneas were discarded after the first analysis due to ECD less than 2000 cells/mm². Thirty-seven others had between 2001 and 2354 cells/mm², but could not be allocated quickly, and their ECDs were deemed too close to the threshold to be allocated after 15 days of storage (borderline ECD). Ten other corneas with ECD more than 2000 cells/mm² were discarded because of endothelial anomalies—most commonly, very high pleomorphism and, more rarely, half of endothelial surface blue after staining with trypan blue; numerous folds; and subjacent stroma remaining blue, suggesting endothelial necrosis. Their endothelial characteristics are shown in Table 5. Those of the 37 borderline corneas were compared to the 248 grafted corneas (Table 6). Initial morphometry of corneas that had borderline ECDs and were finally not grafted was not as good as that of the grafted corneas. Corneas were grafted after 15 days (range, 6–29) of storage. 

Table 5.

 

Endothelial Characteristics of Corneas Eliminated because of Deficient or Borderline Endothelium

	Initial ECD <2000 (n = 68)					ECD Borderline (n = 37)					Other Endothelial Cause (n = 10)
	Mean	SD	Min	Max	Median	Mean	SD	Min	Max	Median	Mean	SD	Min	Max	Median
n	285	57	105	362	304	310	26	232	395	310	301	55	147	334	317
ECD	1499	373	677	1993	1599	2152	97	2002	2354	2162	2677	279	2357	3203	2567
CV	32.2	5.5	21.6	44.1	31.9	33.8	7.3	21.5	48.4	32.6	38.4	10.1	26.3	60.6	35.5
Pentagons, %	28	5.9	18.1	58.8	27.3	26.8	4.1	18.5	33.3	27.2	27.8	4.8	20.9	35	26.8
Hexagons, %	45.8	8.8	17.6	65	45.1	48.9	8.9	30.3	67.3	46.9	47.2	8.8	36.3	64.1	45.6
Heptagons, %	17.8	4	5.8	27.5	18	17.9	4.3	10	28.7	17.5	17.2	4.5	9.8	23.7	17.5
Mortality, %	0.37	0.69	0	5	0.21	0.27	0.6	0	3.3	0.1	0.53	0.66	0.06	1.6	0.17

  n = 105.

Table 6.

 

Comparison of the Initial Analyses of 37 Corneas with Borderline ECD with Those of the 248 Grafted Corneas

	Mean	SD	Minimum	Maximum	Median	P
Cells, n
Discarded	310	26	232	395	310	<0.001
Grafted	330	28	161	417	322
ECD, cells/mm2
Discarded	2152	97	2002	2354	2162	<0.001
Grafted	3112	428	2204	4281	3083
CV, %
Discarded	33.8	7.3	21.5	48.4	32.6	<0.001
Grafted	27.0	4.4	16.4	40.7	26.6
Pentagons, %
Discarded	26.8	4.1	18.5	33.3	27.2	<0.001
Grafted	24.1	4.5	10.7	39.5	23.9
Hexagons, %
Discarded	48.9	8.9	30.3	67.3	46.9	<0.001
Grafted	55.7	7.9	38.7	80.6	56.0
Heptagons, %
Discarded	17.9	4.3	10.0	28.7	17.5	0.079
Grafted	16.5	3.3	7.0	26.6	16.6
Mortality, %
Discarded	0.27	0.60	0.00	3.30	0.10	0.958
Grafted	0.24	0.37	0.00	3.10	0.11

 Mann-Whitney nonparametric test. Discarded, n = 37; grafted, n = 248.

Relations between Image Quality and the Results of Endothelial Analysis

For the 762 endothelial analyses (504 initial + 258 final), 98 (13%) image sets were graded as poor quality, 500 (66%) as average, and 164 (21%) as good. There were significant differences in all parameters, showing deterioration when image quality deteriorated: ECD and hexagonality decreased, and CV and pentagon and heptagon percentages increased (Table 7). There was a nonsignificant trend toward deterioration in image quality with donor age. For the 258 paired endothelial analyses, there was a deterioration in image quality between the start and end of storage, moving from a poor/average/good breakdown of 14 (5%)/142 (55%)/102 (40%) to 35 (13.5%)/195 (75.5%)/28 (11%) (P < 0.01). 

Table 7.

 

Endothelial Analysis Parameters According to Image Quality, in All Analyses (Start and End of OC)

Parameter/Quality of Images	Mean	SD	Min	Max	Median	P
Cells, n
Poor	294	48	105	360	307	<0.01
Average	316	27	116	408	315
Good	338	28	300	420	333
ECD (cells/mm2)
Poor	2386	675	801	3697	2453	<0.01
Average	2645	606	677	4280	2684
Good	2971	561	1291	4281	3001
CV, %
Poor	31.0	6.2	20.1	49.2	30.4	<0.01
Average	28.4	5.7	18.1	60.6	27.3
Good	25.5	4.3	16.4	41.3	24.8
Pentagons, %
Poor	28.7	6.1	12.8	58.8	28.4	<0.01
Average	25.9	4.2	13.4	39.5	25.9
Good	23.5	4.1	10.7	34.8	23.4
Hexagons, %
Poor	46.5	8.6	17.6	67.7	45.3	<0.01
Average	51.2	7.7	30.6	74.6	50.9
Good	57.1	7.0	40.6	80.6	57.2
Heptagons, %
Poor	17.7	4.4	5.8	33.3	17.8	0.008
Average	17.4	3.5	5.8	28.4	17.6
Good	16.5	3.3	5.2	26.6	16.4
Donor age, y
Poor	74	14	36	99	77	0.119
Average	71	15	16	100	73
Good	70	15	34	96	70

  n = 762.

Discussion

We used an image-analysis tool specifically developed to measure ECD and the morphometric parameters of the human corneal endothelium during OC storage and observed them by transmitted light microscopy. ⁶ Compared with specular microscopy images, those obtained by transmitted light microscopy are considered harder to analyze because the cell borders are less visible and background noise is higher. Further, the use of sodium chloride to induce reversible osmotic dilation of the intercellular spaces can modify cell aspect. However, this dilation does not in theory modify neighboring cell relations or size relations between cells. The accuracy ⁸ and reproducibility ²² of the analyses conducted by the analyzer used in the present study have been reported. Analysis comprised cell border recognition and manual correction of poorly recognized boundaries, which is the most reliable method for measuring morphometric parameters. ²⁴ As well as ECD, CV, and hexagonality, the analyzer provides percentages of pentagonal and heptagonal cells, which appear to be important items, at least for research on the relation between morphology and endothelial function. ^13,14 The present work, with a series of 505 corneas routinely analyzed at the eye bank, demonstrates the device's utility for assessing endothelial morphometry despite often difficult imaging conditions. 

For the analysis performed at the start of OC, on average 4 days after donor death, (i.e., closest to physiological state), the measured parameters can be laid over those measured by different authors, in vivo or ex vivo, using different techniques. These are shown in Table 8. Note that our mean ECD tended to be higher than in most studies with specular microscopes. This difference may be due to underestimation by some devices and/or a difference in counting strategy, as we demonstrated for the SP2000P SM (Topcon, Tokyo, Japan) relative to transmitted light microscopy analysis by Sambacornea (Tribvn). ²⁵  

Table 8.

 

Comparison of Ex Vivo Results of Sambacornea with Those in Other Studies

	Yee et al. 26	Chu et al. 27	Müller et al. 28	Reykjavik Eye Study Zoega et al. 29	Cornea Donor Study Sugar et al. 30	Present Study
Year	1985	1995	2004	2006	2006	2009
Technique	In vivo specular	Ex vivo specular	In vivo confocal	In vivo specular	Ex vivo specular	Ex vivo optical
Type of analysis	Manual digitalization	Manual digitalization	Manually assisted count of three images	Not specified; specular software	Computer-aided manual on one image	Semiautomatic on three images
Analyzer	Heye Schule or Keeler Konan	—	Nidek Confoscan 2.0/NAVIS	Noncon ROBO, Konan	All types of specular	Sambacornea
Age, y	70–79	>66	75.7 (10.9)	68 (55–92)	>71	72 (15)
Patients, N	9	895	75	672	168	504
Cells, N	100	—	77–131*	—	50–150	318
ECD, cells/mm2	2630 (60)	2445	2488 (301)	2495 (15)	2692 (273)	2672 (721)
CV, %	29 (1.2)	—	—	36.2 (0.3)	—	29.4 (6.3)
Pentagons, %	19.1 (0.7)	—	—	—	—	25.5 (4.9)
Hexagons, %	60.6 (1.4)	—	47.0 (6.1)†	58.4 (0.4)	—	52.4 (8.9)
Heptagons, %	18.7 (1.1)	—	—	—	—	17.0 (3.6)

 Yee et al., in vivo specular microscopy; Chu et al., ex vivo specular microscopy; Müller et al., in vivo confocal microscopy; Reykjavik Eye Study, in vivo specular microscopy; and Sugar et al, ex vivo specular microscopy at start of storage at +4°C. Results are expressed as the mean (SD). —, no quantitative morphometric data available.

*  Data item extrapolated from the area of interest and ECD described in the article.

†  Lower than usual data attributed by the authors to the small sample size.

The image-analysis system used in this study and using transmitted light microscopy can highlight the same changes in ECD and in morphometric parameters according to age as those already reported in the literature using in vivo and ex vivo SM. ^30–32 However, our population was substantially older than that in the CDS, ³⁰ which explains the difference in the ECD–age relations found in the two studies: biexponential decrease in the CDS, which enrolled more young donors, and linear decrease in our study. Carlson et al. ³² found, in a population of 80 subjects aged 5 to 79 years, an increase in CV of 28% (r = 0.53, P < 0.001), a decrease in the hexagonality of 14% (r = −0.48, P < 0.001), and an increase in percentages of pentagons of 50% (r = 0.44, P < 0.001) and of heptagons of 40% (r = 0.33, P < 0.002). ³² In that study, permeability to fluorescein increased with age. Because the age curve of our population was skewed to the right, we could not calculate these evolutions. Nevertheless, the coefficients of correlation between age and endothelial parameters were on the same order as in Carlson's study, with significant correlations but moderate strengths of association. 

We found a relation between the retriever's subjective assessment and the endothelial morphometric parameters. Although the aged-related bias is obvious, it cannot be excluded that there is a more direct relation between alteration of endothelial parameters and the postmortem corneal condition via increased endothelial permeability, ^11,17,33 which could be responsible for the faster appearance of postmortem edema. 

The Sambacornea analyzer (Tribvn) was also able to highlight morphometric changes caused by cataract surgery, in the same way as Johnston et al. ³⁴ found, in 15 patients with IOL versus 17 phakic control subjects: mean (SD) ECD of 2495 (438) cells/mm² versus 2576 (264) cells/mm² and CV of 32.3% (4%) versus 30.7% (4%), respectively. ³⁴ Age was probably also a factor in the morphologic changes that we found. 

We highlighted a relation between image quality and endothelial parameters, with overall image deterioration when endothelial parameters were less good. This effect seems independent of age. We suppose that image quality reflects the cells' capacity both to react to osmotic preparation (quality of intercellular space dilation) and also to the stromal edema responsible for endothelial folding and image background noise. These highlighted relations tend to show that changes in ECD and in endothelial morphometry result in different corneal behavior and probably reflect an endothelial disorder that is revealed more easily after death in nonphysiological conditions. 

For the subgroup of corneas that had borderline ECD in the first endothelial assessment and were finally not grafted, image analysis also highlighted changed morphometric parameters relative to corneas with higher ECDs. This observation suggests that these corneas were of lower quality and underscores the utility of morphometry, in addition to ECD, for qualifying graft tissue, and therefore probably for aiding in the decision to discard them. The changes highlighted during storage in OC have not been reported. Before this study, only Borderie et al. ^35,36 had reported the influence of the deswelling stage on ECD and morphometry. They highlighted a 27% increase in CV during 48 hours of deswelling and 7% cell loss during 24 hours. We found a mean daily decrease in ECD of 1.38%. We also observed a CV decrease from 27.1% to 25.7%, which indicates cell uniformization, perhaps due to accelerated mortality of cells with size that departs from the mean and which may be more prone to the stresses of death and OC. The decrease in the hexagonality from 55.7% to 50.7% may be explained by the deformation of residual endothelial cells to fill the spaces vacated by dead cells as they desquamate. OC is a particular nonphysiological condition in which the maintained metabolic activities of cells ³⁷ combined with accelerated mortality ^38,39 cause changes to the mosaic, which tend to compensate for defects. In the living body, slower changes are known and are highlighted by SM. A significant increase in polymegethism, associated with a decrease in hexagonality without an overall change in ECD, has been observed in long-term wearers of contact lenses ^40,41 and in cases of diabetes mellitus. ⁴² A decrease in hexagonality with no change in CV or ECD has been described after repeated exposure to ultraviolet rays (arc welding) ^43,44 and in keratoconus. ⁴⁵ Increased pleomorphism in keratoconus may be linked to prolonged wearing of contact lenses. Repeated exposure to infrared rays (glassblowing) has been associated with a decrease in hexagonality and an increase in CV but also with an increase in ECD, suggesting stimulation of cell division. ⁴⁶ The presence of an anterior chamber lens causes substantial changes, increasing CV and cell elongation and decreasing hexagonality. ⁴⁷ After penetrating keratoplasty, a significant decrease in hexagonality without a substantial change in CV indicates that cell enlargement is very gradual. ⁴⁸ These morphologic alterations do not revert to the preoperative level even after 5 years, because of continued cell loss. ⁴⁹  

In conclusion, the Sambacornea (Tribvn) analyzer of the endothelium of corneas stored in OC and observed with transmitted light microscopy produced quantitative morphometric parameters that conformed to those determined in several other studies with SM, despite sometimes difficult imaging conditions and the use of osmotic dilation of the intercellular spaces to visualize the cells. This work concludes a series of studies validating the various aspects of this analyzer. ^6,8,22,24,25 Use of this device should therefore make it possible, in the course of controlled clinical trials, to determine the actual importance of endothelial morphometry in graft quality and survival. 

Footnotes

 Supported by Grant “Recherche et Greffes 2004” from l'Etablissement Français des Greffes, Agence de la Biomédecine.

Footnotes

 Disclosure: S. Acquart, None; P. Gain, None; M. Zhao, None; Y. Gavet, None; A. Defreyn, None; S. Piselli, None; O. Garraud, None; G. Thuret, None

The authors thank Sandrine Pereira and Christian Theillière for active participation in all the endothelial analyses.