In vivo confocal microscopy of the corneal endothelium: comparison of three morphometry methods after corneal transplantation

Machine learning based endothelial cell image analysis of patients undergoing descemet membrane endothelial keratoplasty surgery

OBJECTIVES

In this study, we developed a machine learning approach for postoperative corneal endothelial cell images of patients who underwent Descemet's membrane keratoplasty (DMEK).

METHODS

An AlexNet model is proposed and validated throughout the study for endothelial cell segmentation and cell location determination. The 506 images of postoperative corneal endothelial cells were analyzed. Endothelial cell detection, segmentation, and determining of its polygonal structure were identified. The proposed model is based on the training of an R-CNN to locate endothelial cells. Next, by determining the ridges separating adjacent cells, the density and hexagonality rates of DMEK patients are calculated.

RESULTS

The proposed method reached accuracy and F1 score rates of 86.15 % and 0.857, respectively, which indicates that it can reliably replace the manual detection of cells in vivo confocal microscopy (IVCM). The AUC score of 0.764 from the proposed segmentation method suggests a satisfactory outcome.

CONCLUSIONS

A model focused on segmenting endothelial cells can be employed to assess the health of the endothelium in DMEK patients.

Potential applications of artificial intelligence in image analysis in cornea diseases: a review

Artificial intelligence (AI) is an emerging field which could make an intelligent healthcare model a reality and has been garnering traction in the field of medicine, with promising results. There have been recent developments in machine learning and/or deep learning algorithms for applications in ophthalmology-primarily for diabetic retinopathy, and age-related macular degeneration. However, AI research in the field of cornea diseases is relatively new. Algorithms have been described to assist clinicians in diagnosis or detection of cornea conditions such as keratoconus, infectious keratitis and dry eye disease. AI may also be used for segmentation and analysis of cornea imaging or tomography as an adjunctive tool. Despite the potential advantages that these new technologies offer, there are challenges that need to be addressed before they can be integrated into clinical practice. In this review, we aim to summarize current literature and provide an update regarding recent advances in AI technologies pertaining to corneal diseases, and its potential future application, in particular pertaining to image analysis.

Assessing Corneal Endothelial Damage Using Terahertz Time-Domain Spectroscopy and Support Vector Machines

The endothelial layer of the cornea plays a critical role in regulating its hydration by actively controlling fluid intake in the tissue via transporting the excess fluid out to the aqueous humor. A damaged corneal endothelial layer leads to perturbations in tissue hydration and edema, which can impact corneal transparency and visual acuity. We utilized a non-contact terahertz (THz) scanner designed for imaging spherical targets to discriminate between ex vivo corneal samples with intact and damaged endothelial layers. To create varying grades of corneal edema, the intraocular pressures of the whole porcine eye globe samples (n = 19) were increased to either 25, 35 or 45 mmHg for 4 h before returning to normal pressure levels at 15 mmHg for the remaining 4 h. Changes in tissue hydration were assessed by differences in spectral slopes between 0.4 and 0.8 THz. Our results indicate that the THz response of the corneal samples can vary according to the differences in the endothelial cell density, as determined by SEM imaging. We show that this spectroscopic difference is statistically significant and can be used to assess the intactness of the endothelial layer. These results demonstrate that THz can noninvasively assess the corneal endothelium and provide valuable complimentary information for the study and diagnosis of corneal diseases that perturb the tissue hydration.

A Review On digital image processing techniques for in-Vivo confocal images of the cornea

This work reviews the scientific literature regarding digital image processing for in vivo confocal microscopy images of the cornea. We present and discuss a selection of prominent techniques designed for semi- and automatic analysis of four areas of the cornea (epithelium, sub-basal nerve plexus, stroma and endothelium). The main context is image enhancement, detection of structures of interest, and quantification of clinical information. We have found that the preprocessing stage lacks of quantitative studies regarding the quality of the enhanced image, or its effects in subsequent steps of the image processing. Threshold values are widely used in the reviewed methods, although generally, they are selected empirically and manually. The image processing results are evaluated in many cases through comparison with gold standards not widely accepted. It is necessary to standardize values to be quantified in terms of sensitivity and specificity of methods. Most of the reviewed studies do not show an estimation of the computational cost of the image processing. We conclude that reliable, automatic, computer-assisted image analysis of the cornea is still an open issue, constituting an interesting and worthwhile area of research.

Unbiased corneal tissue analysis using Gabor-domain optical coherence microscopy and machine learning for automatic segmentation of corneal endothelial cells

SIGNIFICANCE

An accurate, automated, and unbiased cell counting procedure is needed for tissue selection for corneal transplantation.

AIM

To improve accuracy and reduce bias in endothelial cell density (ECD) quantification by combining Gabor-domain optical coherence microscopy (GDOCM) for three-dimensional, wide field-of-view (1  mm2) corneal imaging and machine learning for automatic delineation of endothelial cell boundaries.

APPROACH

Human corneas stored in viewing chambers were imaged over a wide field-of-view with GDOCM without contacting the specimens. Numerical methods were applied to compensate for the natural curvature of the cornea and produce an image of the flattened endothelium. A convolutional neural network (CNN) was trained to automatically delineate the cell boundaries using 180 manually annotated images from six corneas. Ten additional corneas were imaged with GDOCM and compared with specular microscopy (SM) to determine performance of the combined GDOCM and CNN to achieve automated endothelial counts relative to current procedural standards.

RESULTS

Cells could be imaged over a larger area with GDOCM than SM, and more cells could be delineated via automatic cell segmentation than via manual methods. ECD obtained from automatic cell segmentation of GDOCM images yielded a correlation of 0.94 (p  <  0.001) with the manual segmentation on the same images, and correlation of 0.91 (p  <  0.001) with the corresponding manually counted SM results.

CONCLUSIONS

Automated endothelial cell counting on GDOCM images with large field of view eliminates selection bias and reduces sampling error, which both affect the gold standard of manual counting on SM images.

Comparison of automated vs manual analysis of corneal endothelial cell density and morphology in normal and corneal endothelial dystrophy-affected dogs

OBJECTIVE

To determine the efficacy of automated imaging software of the Nidek ConfoScan 4 confocal biomicroscope at analyzing canine corneal endothelial cell density and morphology in health and disease, by comparing to a manual analysis method.

ANIMAL STUDIED

Nineteen eyes of 10 dogs were evaluated and include three Beagles, three Jack Russell Terriers, and four miscellaneous breeds. Twelve clinically normal and seven eyes affected with corneal endothelial dystrophy (CED) were scanned and analyzed.

PROCEDURES

Endothelial cell density (ECD), mean and standard deviation (SD) of cell area, percent polymegathism, mean and SD of the number of cell sides, and percent pleomorphism were calculated using automated and manual methods for each scan.

RESULTS

The automated analysis showed significantly greater ECD in comparison with the manual frame method due to misidentification of cell domains in CED-affected dogs. No significant differences in ECD were observed between normal and CED-affected dogs in automated analysis, while CED-affected dogs showed significantly lower ECD in manual frame method and planimetry. Using both automated and manual methods, CED-affected dogs showed greater variability of cell area or the number of cell sides than normal dogs.

CONCLUSION

The automated imaging software is unable to accurately identify cell borders in CED-affected dogs resulting in inaccurate estimates of ECD. Thus, manual analysis is recommended for use in clinical trials assessing adverse events associated with novel medical treatments and/or surgical procedures.

Comparison of Noncontact Specular and Confocal Microscopy for Evaluation of Corneal Endothelium

PURPOSE

To compare endothelial cell analysis obtained by noncontact specular and confocal microscopy, using the Konan NSP-9900 and Nidek ConfoScan4 systems, respectively.

METHODS

Three groups including 70 healthy eyes, 49 eyes with Fuchs endothelial corneal dystrophy (FECD), and 78 eyes with glaucoma were examined with both the Konan NSP-9900 specular microscope and the Nidek ConfocScan4 confocal microscope. Certified graders at the Doheny Image Reading Center compared corneal endothelial images from both instruments side by side to assess image quality. Endothelial cell density (ECD) measurements were calculated and compared using three different modalities: (1) each instrument's fully automated analysis; (2) each instrument's semiautomatic analysis with grader input; and (3) manual grading methods by certified grader.

RESULTS

All normal eyes yielded gradable endothelial images, and most but not all glaucomatous eyes yielded images with high enough image quality to allow grading. In addition, in corneas with severe FECD, poor image quality precluded ECD grading by specular microscopy in 20 eyes (40.8%) but in only 4 (8.2%) confocal images from the same eyes. For the gradable images, the ECD values obtained using the manual grading method from either device were comparable with no statistically significant difference (P>0.05) between specular and confocal devices. Machine-generated ECD values were significantly different from manual results, measuring greater in all cases with specular microscopy. Machine-generated ECD values from confocal microscopy also differed significantly from manual determinations, but not in a consistent direction. Semiautomatic methods for both instruments obtained clinically acceptable ECD values.

CONCLUSIONS

Automatic machine-generated ECD measurements differed significantly from manual assessments of corneal endothelium by both specular and confocal microscopy, suggesting that automated results should be used with caution. But ECD values derived manually were comparable between the two devices in both normal and glaucomatous eyes, suggesting that manually graded images from the two instruments can be used interchangeably for reliable ECD measurements. Because of a higher proportion of gradable images, confocal microscopy may be superior to specular microscopy for ECD measurements in FECD.

Corneal Endothelial Cell Segmentation by Classifier-Driven Merging of Oversegmented Images

Corneal endothelium images obtained by in vivo specular microscopy provide important information to assess the health status of the cornea. Estimation of clinical parameters, such as cell density, polymegethism, and pleomorphism, requires accurate cell segmentation. State-of-the-art techniques to automatically segment the endothelium are error-prone when applied to images with low contrast and/or large variation in cell size. Here, we propose an automatic method to segment the endothelium. Starting with an oversegmented image comprised of superpixels obtained from a stochastic watershed segmentation, the proposed method uses intensity and shape information of the superpixels to identify and merge those that constitute a cell, using support vector machines. We evaluated the automatic segmentation on a data set of in vivo specular microscopy images (Topcon SP-1P), obtaining 95.8% correctly merged cells and 2.0% undersegmented cells. We also evaluated the parameter estimation against the results of the vendor's built-in software, obtaining a statistically significant better precision in all parameters and a similar or better accuracy. The parameter estimation was also evaluated on three other data sets from different imaging modalities (confocal microscopy, phase-contrast microscopy, and fluorescence confocal microscopy) and tissue types (ex vivo corneal endothelium and retinal pigment epithelium). In comparison with the estimates of the data sets' authors, we achieved statistically significant better accuracy and precision in all parameters except pleomorphism, where a similar accuracy and precision were obtained.

Anterior segment optical coherence tomography: its application in clinical practice and experimental models of disease

Optical coherence tomography (OCT) provides non-invasive, high-resolution in vivo imaging of the ocular surface and anterior segment. Over the years, it has become an essential tool for evaluating the anterior segment of the eye to monitor ocular development and ocular pathologies in both the clinical and research fields of ophthalmology and optometry. In this review, the clinical applications relating to the use of anterior segment OCT for imaging and quantifying normal and pathological features of the ocular surface, cornea, anterior chamber, and aqueous outflow system are summarised in a range of human ocular diseases. Applications of anterior segment OCT technology that have improved imaging and quantitation of ocular inflammation in experimental animal models of ocular diseases, such as anterior uveitis, microbial keratitis and glaucoma, are also described. The capacity to longitudinally evaluate anterior segment anatomical changes during development, and inflammation facilitates the understanding of the dynamics of tissue responses, and further enhances the intra-operative in vivo imaging during procedures, such as corneal transplantation and drug delivery. Future developments including in vivo ultrahigh-resolution anterior segment OCT, automated analyses of anterior segment OCT images and functional extensions of the technique, may revolutionise the clinical evaluation of anterior segment, corneal and ocular surface diseases.

Peripheral Endothelial Cell Count Is a Predictor of Disease Severity in Advanced Fuchs Endothelial Corneal Dystrophy

PURPOSE

In advanced Fuchs endothelial corneal dystrophy (FECD), central endothelial changes do not correlate with disease severity. The peripheral endothelial cell count (ECC) has not been studied as a marker of FECD severity. The goal of this study was to determine the relationship between the peripheral ECC and known clinical markers of FECD in advanced cases.

METHODS

Patients with FECD examined between January 1, 2013, and September 1, 2016, by 1 cornea specialist were identified. Medical records from all previous visits were reviewed to include eyes with high-quality central and peripheral in vivo confocal microscopy images performed on the same day as a clinical evaluation. Endothelial photographs were used to perform manual cell counts centrally and peripherally. Clinical grading of FECD from 1 to 4 was performed at the slit-lamp.

RESULTS

We identified 154 eyes of 126 patients that met criteria for inclusion. With higher disease grades, central ECC and peripheral ECC decreased, visual acuity worsened, and central corneal thickness (CCT) increased (all P < 0.05). In patients with advanced disease (defined as either grade 3 or 4, CCT >700, or central ECC <350), the peripheral ECC was the best predictor of disease severity and had the highest number of statistically significant correlations with other clinical markers compared with competing variables.

CONCLUSIONS

In advanced FECD, severity is best determined by the peripheral ECC compared with the central ECC, visual acuity, clinical disease grade, and CCT. The peripheral ECC should be added to the clinical parameters used to evaluate FECD severity.

Smartphone-based imaging of the corneal endothelium at sub-cellular resolution

This aim of this study was to test the feasibility of smartphone-based specular microscopy of the corneal endothelium at a sub-cellular resolution. Quantitative examination of endothelial cells is essential for evaluating corneal disease such as determining a diagnosis, monitoring progression and assessing treatment. Smartphone-based technology promises a new opportunity to develop affordable devices to foster quantitative examination of endothelial cells in rural and underserved areas. In our study, we incorporated an iPhone 6 and a slit lamp to demonstrate the feasibility of smartphone-based microscopy of the corneal endothelium at a sub-cellular resolution. The sub-cellular resolution images allowed quantitative calculation of the endothelial cell density. Comparative measurements revealed a normal endothelial cell density of 2978 cells/mm² in the healthy cornea, and a significantly reduced cell density of 1466 cells/mm² in the diseased cornea with Fuchs' dystrophy. Our ultimate goal is to develop a smartphone-based telemedicine device for low-cost examination of the corneal endothelium, which can benefit patients in rural areas and underdeveloped countries to reduce health care disparities.

Fully automatic evaluation of the corneal endothelium from in vivo confocal microscopy

Background

Manual and semi-automatic analyses of images, acquired in vivo by confocal microscopy, are often used to determine the quality of corneal endothelium in the human eye. These procedures are highly time consuming. Here, we present two fully automatic methods to analyze and quantify corneal endothelium imaged by in vivo white light slit-scanning confocal microscopy.

Methods

In the first approach, endothelial cell density is estimated with the help of spatial frequency analysis. We evaluate published methods, and propose a new, parameter-free method. In the second approach, based on the stochastic watershed, cells are automatically segmented and the result is used to estimate cell density, polymegathism (cell size variability) and pleomorphism (cell shape variation). We show how to determine optimal values for the three parameters of this algorithm, and compare its results to a semi-automatic delineation by a trained observer.

Results

The frequency analysis method proposed here is more precise than any published method. The segmentation method outperforms the fully automatic method in the NAVIS software (Nidek Technologies Srl, Padova, Italy), which significantly overestimates the number of cells for cell densities below approximately 1200 mm⁻², as well as previously published methods.

Conclusions

The methods presented here provide a significant improvement over the state of the art, and make in vivo, automated assessment of corneal endothelium more accessible. The segmentation method proposed paves the way to many possible new morphometric parameters, which can quickly and precisely be determined from the segmented image.

Electronic supplementary material

The online version of this article (doi:10.1186/s12880-015-0054-3) contains supplementary material, which is available to authorized users.

Objective assessment of changes in nuclear morphology and cell distribution following induction of apoptosis

BACKGROUND

To objectively measure changes in nuclear morphology and cell distribution following induction of apoptosis.

METHODS

A spontaneously immortalized retinal pigment epithelial cell line (ARPE-19) was cultured for three days in DMEM/F12 with 10% fetal bovine serum followed by 24 hours incubation in staurosporine to induce apoptosis. Cells that were not incubated in staurosporine served as control. Caspase-3 expression in apoptotic cells was demonstrated by quantitative immunofluorescence. Nuclei were counterstained with DAPI. Assessments of nuclear morphology and cell distribution were performed using ImageJ software. Statistical analyses included Student's t-test and Pearson's correlation coefficient. Nearest neighbor analysis was used to assess cell nuclei distribution.

RESULTS

Caspase-3 expression in staurosporine-incubated cells increased by 471% ± 182% compared to control (P=0.014). Relative to the control, cells in the staurosporine-incubated cultures had smaller average nuclear area (68% ± 5%; P<0.001) and nuclear circumference (78 ± 3%; P<0.001), while nuclear form factor was larger (110% ± 1%; P<0.001). Cell nuclei from the staurosporine-group (R=1.12 ± 0.04; P<0.01) and the control (R=1.28 ± 0.03; P<0.01) were evenly spaced throughout the cultures, thereby demonstrating a non-clustered and non-random cell distribution. However, the staurosporine-incubated group had a significantly lower R-value compared to the control (P=0.002), which indicated a move towards cell clustering following induction of apoptosis. Caspase-3 expression of each individual cell correlated significantly with the following morphological indicators: circumference of the nucleus divided by form factor (r=-0.475; P<0.001), nuclear area divided by form factor (r=-0.470; P<0.001), nuclear circumference (r=-0.469; P<0.001), nuclear area (r=-0.445; P<0.001), nuclear form factor (r=0.410; P<0.001) and the nuclear area multiplied by form factor) (r=-0.377; P<0.001).

CONCLUSIONS

Caspase-3 positive apoptotic cells demonstrate morphological features that can be objectively quantified using freely available ImageJ software. A novel morphological indicator, defined as the nuclear circumference divided by form factor, demonstrated the strongest correlation with caspase-3 expression.

VIRTUAL SLIDES

The virtual slide(s) for this article can be found here: http://www.diagnosticpathology.diagnomx.eu/vs/3271993311662947.

Objective assessment of the corneal endothelium in Fuchs' endothelial dystrophy

PURPOSE

To develop a standardized method of endothelial cell density (ECD) assessment in Fuchs' endothelial dystrophy that maximizes the sample area and uses the clearest endothelial cells in confocal images.

METHODS

The corneal endothelium of 51 eyes from 30 patients, with varying degrees of Fuchs' endothelial dystrophy, was examined using confocal microscopy. In two or three distinct images of the central endothelium, local contiguous cell density was determined using a variable frame method. The effective ECD was the product of the local cell density and the fraction of the image that was free of guttae. Two examiners assessed the severity of disease in each eye during slit-lamp examination and assigned a severity grade of 1 to 6. In a second group of 55 eyes with Fuchs' dystrophy from 30 patients, the clinical grade was predicted from the effective ECD and the regression coefficients of the first group and compared to the subjective clinical grade assigned by one examiner.

RESULTS

The effective ECD decreased linearly with subjective grade (r = -0.93, P < 0.001). The grade predicted from the effective ECD differed from the subjective clinical grade by -0.1 ± 0.8 (mean difference ± standard deviation).

CONCLUSIONS

The effective ECD in confocal images provides an objective means of assessing the corneal endothelium in Fuchs' dystrophy and might be a useful tool in clinical studies.

Assessment of central corneal thickness and corneal endothelial morphology using ultrasound pachymetry, non-contact specular microscopy, and Confoscan 4 confocal microscopy

BACKGROUND

The aim was to compare the repeatability, reproducibility and inherent precision of ultrasound pachymetry (USP), noncontact specular microscopy (SP-2000P) and the Confoscan 4 confocal microscope (z-ring CS4) in measuring endothelial cell density (ECD), coefficient of variation of cell size (CV), and central corneal thickness (CCT) in normal eyes.

METHODS

In this prospective study, one eye was selected from each of 30 subjects for the measurements of ECD, CV and CCT, which were taken by two observers. Results were analyzed statistically by repeated-measures analysis of variance (ANOVA) for intra-observer repeatability, inter-observer reproducibility, unpaired t-test, paired t-test, and Bland-Altman analyses to determine limits of agreement (LOA) between the three instruments.

RESULTS

Mean ECD, measured by SP-2000P and z-ring CS4, were 3115.50 ± 279.70 cells/mm(2) and 3167.50 ± 264.75 cells/mm(2), respectively (observer 1), and 3192.63 ± 249.42 cells/mm(2) (z-ring, observer 2). Mean CV measurements were 27.12 ± 2.51 and 27.10 ± 2.41 (SP-2000P and z-ring CS4, respectively; observer 1), and 27.17 ± 2.25 (z-ring, observer 2). Mean CCT values were 555.11 ± 35.83 μm (USP), 535.82 ± 41.10 μm (SP-2000P) and 552.57 ± 36.83 μm (z-ring CS4), and 554.97 ± 36.34 μm (z-ring CS4, observer 2). However, pairwise tests in all cases there was good repeatability and reproducibility as shown by inter-observer and intra-observer analysis of variance for each of the instruments.

CONCLUSIONS

The SP-2000P and the z-ring CS4 can be used interchangeably to measure ECD and CV. For CCT, the sample size was too small to test for differences of the CCT measurements between the three instruments.

Assessment of a variable frame (polygonal) method to estimate corneal endothelial cell counts after corneal transplantation

PURPOSE

To assess the agreement of the 'polygonal' variable frame cell count option on a confocal microscope after keratoplasty, with planimetry as the reference method.

METHODS

One hundred clear corneal grafts of 83 patients attending the cornea clinic at Gartnavel General Hospital in Glasgow underwent slit-scanning in vivoconfocal microscopy. Endothelial cell images were assessed with the Nidek Advanced Vision Information System (NAVIS), using the polygonal variable frame and the manual fixed-frame methods. Planimetry was used as the reference. The agreement between methods was assessed by Bland-Altman analysis.

RESULTS

Planimetry provided a mean (± SD) endothelial cell density (ECD) of 1348 ± 726 cells/mm(2), a value that was very similar to that found by the polygonal method (1404 ± 784 cells/mm(2)). The fixed-frame method provided lower cell counts with a mean ECD of 1026 ± 610 cells/mm(2) (P<0.001). When compared with the reference ECD, the polygonal method overestimated the ECD only very slightly with a mean difference of 58 cells/mm(2) (limits of agreement, LoA, of -222 and 339 cells/mm(2)). Manual counting underestimated the ECD with a mean difference of -320 cells/mm(2) (LoA -814 and 173 cell/mm(2)).

CONCLUSION

Following keratoplasty, endothelial cell counts with the NAVIS polygonal method are in good agreement with planimetry. The 'polygonal' option is proposed as the method of choice for clinical applications with this confocal microscope and a good compromise between reliability and ease of use.

Assessment of the reliability of endothelial cell-density estimates in the presence of pseudoguttata

PURPOSE

The purpose of this work is to assess the reliability of endothelial cell-density (ECD) estimates in corneas with different severity pseudoguttata.

METHODS

Specular microscopy was undertaken on grade 1, 2, or 3 pseudoguttata patients and age-matched controls aged 52-83 years. On high magnification prints of central cornea, areas of complete cells (all sides visible) and partial 'cells' (one or more sides obscured) were measured manually. Sets of 45 complete cells were selected, as well as 75 cells that were a mixture of complete and partial cells on guttate endothelia. ECD was calculated by a progressive averaging technique.

RESULTS

Each group comprised 12 patients with similar range of ECD values (1,230-4,587 cells/mm(2)). Based on 40 complete cells, ECD could be estimated to within ±3.1% for grade 3 pseudoguttata versus ±2.0% for controls. If a mixture of complete and partial cells were measured, ECD could be estimated to within ±2.8% for grade 3 pseudoguttata images (n = 70 cells) and ±1.1% for controls. The estimated variability increases to substantial levels of ±20% if only ten cells were measured. No statistical differences in ECD were noted between guttate and normal endothelia if only complete cells were measured, but could be different if partial 'cells' were included.

CONCLUSIONS

Providing adequate numbers of complete cells are measured and in the absence of obvious polymegathism, ECD estimates can be made to within around ±3% in the presence of typical but significant pseudoguttata.