In Vitro Corneal Tomography of Donor Cornea Using Anterior Segment OCT

BASIC INVESTIGATION

Edgar Janunts, PhD,* Achim Langenbucher, PhD,* and Berthold Seitz, ML, MD†

n cases of corneal donation, either the entire globe or just Purpose: The aim of this study was to establish a tomographic Ia 15-mm corneoscleral piece of tissue is taken. Having screening method for revealing potential pathologies in corneal confirmed the absence of obvious scars on the cornea and donors before keratoplasty so they may be excluded as candidates for a medical history without any corneal surgeries or infectious corneal transplantation. diseases, the corneoscleral button can be excised and stored in fi “ ” Methods: Donor corneal tomographies were measured in a viewing a special chamber lled with a transport medium before chamber filled with preservation medium and with the use of transplantation. At present, only a visual examination using a clinical optical coherence tomography (OCT) device. Custom- a handheld slit lamp is performed before the corneoscleral written software was developed to extract corneal surfaces from the excision, in addition to blood and smear tests. Further slit- raw data, which were analyzed in the central and peripheral regions. lamp examinations and microbiological tests are usually An adaptive nonlinear edge-enhancement algorithm was used to performed under laboratory conditions in eye banks. There observe scars within the corneal volume. The thickness distribution is no tomographic screening routine presently available that map was analyzed to detect keratoconus and corneas with extreme discounts corneas that might be unsuitable for future trans- topographic irregularities. Measurements were repeated 5 times to plant, such as those which have suffered previous surgeries assess reproducibility. (eg, LASIK) or diseases (eg, keratoconus) that might affect the biomechanical properties of the cornea. Results: Eight corneas were investigated: 6 randomly selected There are a few case reports in the literature in which, intact donors, unsuitable for implantation because of low endothelial despite the routine examinations mentioned above, donated cell densities, and 2 keratoconus corneas, excised from patients corneas, which had previously undergone an operation, were during corneal transplantation. A major thickness abnormality was accepted for transplantation. Two clinical cases were reported1 detected in one of the intact donor corneas, so it was excluded from where corneas were only diagnosed as having undergone further analysis. The keratoconus corneas were clearly evident in a LASIK procedure after they had been transplanted. Ehren- optical coherence tomography cross-sectional images, and similarly, haus et al2 reported a case of a donor cornea with a soft contact fi they could easily be identi ed by analyzing the thickness map. lens still in place. A recent study by Kang et al3 estimated that ’ Overall, the measurements were reliable and had a Cronbach s alpha approximately 10% of corneas in eye banks have undergone an fi coef cient greater than 0.8. as yet unidentified LASIK procedure. – Conclusions: Donor corneal examination using sterile viewing In the past, Placido ring based video keratographs were 4,5 chambers was found to be suitable as a pre-keratoplasty advanced adapted for use under eye bank conditions, whereas more screening routine. A proof of concept was demonstrated, which recent studies favor the screening of donor corneal tissue with 6,7 could identify both irregular corneas and those affected the use of optical coherence tomography (OCT). Ousley 4 by keratoconus. and Terry introduced a portable, Placido-based topography instrument, known as a Keratron Scout, for donor screening. Key Words: eye bank, donor corneal examination, keratoconus, The authors also proposed a method for the identification of keratoplasty, tomographic screening tissues that had undergone refractive surgery by observing the (Cornea 2016;35:647–653) difference in corneal curvature at the 7- and 3-mm zones. They showed that corneal curvature does not change much before and after excision from the donor; however, moderate changes in astigmatism could be observed.4 Lin8 developed a laboratory OCT system to perform Received for publication August 4, 2015; revision received November 23, corneal screening in a viewing chamber; the authors used 2015; accepted December 13, 2015. Published online ahead of print a custom-made high-resolution laser source and monitored March 1, 2016. From the *Institute of Experimental Ophthalmology, University of Saarland, refractive index distributions to differentiate between normal Homburg, Saar, Germany; and †Department of Ophthalmology, Saarland corneas and LASIK-operated corneas. In a different work, Lin University Medical Center, Homburg, Saar, Germany. et al9 proposed a method to detect LASIK-operated corneas, The authors have no funding or conflicts of interest to disclose. which involved measuring the anterior curvature and the Reprints: Edgar Janunts, PhD, Institute of Experimental Ophthalmology, fl Saarland University, Kirrberger St 100, Bldg 22, 66421 Homburg/Saar, anterior/posterior stromal re ectivity ratio with OCT scans. Germany (e-mail: [email protected]). The present study aims to propose a tomographic screening Copyright © 2016 Wolters Kluwer Health, Inc. All rights reserved. method to detect possible pathologies in donor material and

Cornea Volume 35, Number 5, May 2016 www.corneajrnl.com | 647

Copyright Ó 2016 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. Janunts et al Cornea Volume 35, Number 5, May 2016 exclude these potential donors from future transplantations data processing. The image processing software automatically (among others, corneas that have undergone refractive detected the cornea and corrected the images by compensat- surgery and those with geometrical abnormalities, ing the refraction error of the scanning beam passing through eg, keratoconus). layers with different refractive indices [air, cuvette (glass), liquid, cornea, and liquid].

MATERIALS AND METHODS Image Processing Algorithm The donor corneal examination was performed using The contrast of the OCT images was enhanced at first, a viewing chamber from Bausch and Lomb by an AS-OCT then the images were smoothed using a standard Gaussian SS-1000 CASIA clinical tomographer (Tomey) in a laboratory filter, and the noise was removed using LabVIEW’s particle environment without removing the corneas from the sterile analysis algorithm. The corneal surfaces were then extracted preservation medium (Fig. 1). with Canny edge filtering and presented as separate surfaces. The viewing chamber was filled with transport medium, Finally, the surfaces were corrected according to Snell’s law a commercial product from Biochrom that consists of the (Fig. 2), and the thickness map was computed. following substances: MEM Earle’s salt (·10), penicillin/ streptomycin, L-glutamine (200 mM), amphotericin B (250 mg/mL), HEPES buffer (1 M), and NaHCO3. Methods of Analysis After measuring the cornea, a custom-written software Based on the anterior surface and the pachymetry maps using LabVIEW 2013 (National Instruments) was applied for (Figs. 3, 4), various automatic analyses were programmed to

FIGURE 1. A 3-dimensional cross-sectional examination of donor corneoscleral tissue using an SS-1000 CASIA clinical OCT tomographer (Tomey).

Copyright Ó 2016 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. Cornea Volume 35, Number 5, May 2016 Donor Corneal Tomography Using Anterior Segment OCT

FIGURE 2. Image processing algorithm. The contrast of the raw image [OCT cross-sectional scan (A)] was enhanced and a standard Gaussian filter applied (B). Particle analysis was then carried out to reduce noise, thus removing all particles unrelated to the cornea (C). In subsequent steps, the corneal volume was separated by applying Canny edge filters (D). Finally, both corneal surfaces were recognized (E) and extracted as separate curves (F). identify donor corneas that may potentially be unacceptable Furthermore, in addition to the adaptive edge-enhanced for keratoplasty. approach, a mathematical evaluation model was used to In particular, a mathematical model was used to diagnose the presence of postrefractive surgery corneas by diagnose scars in the corneal tissue, such as flaps resulting comparing the difference in anterior corneal curvature of the from LASIK surgery, intracorneal rings, or femtosecond central (3 mm) and peripheral (7 mm) regions, as proposed by laser–assisted keratoplasties, by using an adaptive edge- Ousley and Terry.4 Both spherical and aspherical parameters enhancement algorithm. Here a color-coded cross-sectional were analyzed. image was computed (Fig. 3), where brighter coloring To detect corneas with keratoconus disease or thickness indicated a higher contrast. The algorithm comprises a local abnormalities, analysis of entire thickness distribution of the filtering method based on a Taylor expansion.10 The filtered cornea was performed by superimposing all cross-sectional images were smoothed using Gaussian smoothing followed pachymetries (B-scans) onto a 2-dimensional graph, as shown by a nonlinear filter for detecting local discontinuities. in Figs. 4–7. This new approach defines the thickness of the cornea in the center and then the maximum difference in thickness with respect to the center for each ring profile.

Statistical Analysis Five complete measurements were performed and analyzed to assess the reproducibility. Statistical analysis was performed with the SPSS version 20.0 statistical software package (SPSS). All data were analyzed and presented with their mean, median, and SD values. Cronbach’s alpha coefﬁcient11,12 was calculated as a measure of reliability.

FIGURE 3. A cross-sectional volume map of a donor cornea RESULTS imaged in a viewing chamber. The corneal volume map was Eight corneas were investigated: 6 randomly selected obtained after analysis of a raw image by applying an adaptive intact donor corneas, unsuitable for implantation because of edge-enhancement algorithm. their low endothelial cell densities, and 2 corneas excised

Copyright Ó 2016 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. Janunts et al Cornea Volume 35, Number 5, May 2016

FIGURE 4. Custom-written software for automatic analysis of donor cornea. On the left: anterior cornea overlaid with color-coded pachymetry; on the right: cross-sectional thickness distributions. from keratoconus patients undergoing corneal transplantation. Three exemplary cases are presented graphically: A thickness abnormality was detected in one of the intact a normal cornea (Fig. 5), a keratoconus cornea excised from a donor corneas, so it was excluded from the parametric patient undergoing corneal transplantation (Fig. 6), and analysis. The results of analyses are summarized in Table 1. a cornea with an abnormal thickness distribution (Fig. 7).

TABLE 1. Parametric Analyses of Anterior Corneal Surfaces and Pachymetry Data for donor Corneas Are Shown Anterior Cornea 3-mm Area 7-mm Area Pachymetry Donor Cornea Sphere: ROC (mm) Asphere: ROC (mm) Sphere: ROC (mm) Asphere: ROC (mm) Central (mm) Thinnest Point (mm) 1 Mean 6 SD 7.6 6 0.04 7.7 6 0.03 7.8 6 0.03 7.7 6 0.02 650 6 9.3 646 6 8 Median 7.6 7.7 7.8 7.7 653 647 2 Mean 6 SD 7.4 6 0.1 7.6 6 0.1 7.6 6 0.03 7.6 6 0.1 675 6 5.9 668 6 2.7 Median 7.5 7.5 7.6 7.5 674 667 3 Mean 6 SD 7.6 6 0.09 7.8 6 0.2 7.7 6 0.07 7.8 6 0.2 695 6 38.8 690 6 35.7 Median 7.7 7.9 7.8 7.9 709 702 4 Mean 6 SD 7.4 6 0.2 7.7 6 0.2 7.9 6 0.07 7.7 6 0.1 703 6 7.1 679 6 6.3 Median 7.37 7.74 7.76 7.69 705 674 5 Mean 6 SD 7.8 6 0.06 8 6 0.2 7.9 6 0.06 8 6 0.16 639 6 3.5 634 6 5 Median 7.8 8 7.9 8 640 635 Cronbach’s a 0.832 0.826 0.955 0.871 0.916 0.887

Two different areas of anterior corneal surface were analyzed at 3 and 7 mm, and the thickness in the center and at the thinnest point is given. ROC refers to a spherical radius of curvature. Each cornea was measured 5 times; hence, the mean, SD, and the median for each cornea were calculated as was the Cronbach’s alpha coefﬁcient for all parameters.

Copyright Ó 2016 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. Cornea Volume 35, Number 5, May 2016 Donor Corneal Tomography Using Anterior Segment OCT

FIGURE 5. Tomographic analysis of a normal donor cornea. A 3-dimensional OCT scan was performed with a CASIA corneal tomographer (Tomey) (A). A pachymetry distribution map (B) was generated by calculating a thickness proﬁle for each cross-sectional OCT image.

DISCUSSION topographers because they enable measurements from a ver- The detection of previous refractive surgeries or corneal tical position before excision of the corneoscleral button or diseases in potential donor corneas before performing kera- even enucleation, but they are not suitable for advanced donor toplasty is a challenge that corneal transplant surgery corneal screening within sterile conditions. presently faces. It is almost impossible to detect keratoconus We have made additional efforts to scan donor corneas in donor corneas with the current examination techniques using laser scanning confocal sensors with submicron reso- available in eye banks. lution, the kind commonly used in optical metrology (optical Significant irregularities and optical aberrations are often imaging) to analyze multilayer transparent objects. However, noticed after penetrating keratoplasty. In most cases, these despite its ability to analyze 3-dimensional surfaces, for irregularities are associated with the surgical procedure. unknown reasons it could not complete a reasonable mea- However, to some degree, they are also associated with surement of the entire donor cornea in a conventional cell preexisting irregularities present in the donor cornea. Analysis culture flask filled with preservation medium, one of the main of donor corneas before and after transplantation could identify reasons being the high dynamic range of incident angles; the whether the astigmatism was truly introduced by the surgery. angle between the viewing direction (vertical) and the surface Portable Placido ring–based donor corneal topogra- normal to the cornea. The greater the angle, the lesser the phers (eg, the “Keratron Scout” device4) have been designed intensity of the light reflected to the scanner. Under optimal for screening donor corneas in eye banks. However, these conditions, these sensors operate up to the seventh deviation portable topographers have been shown to be unsuitable and of surface, which corresponds to a central corneal area with insufficient for screening purposes before keratoplasty. The 2 a maximum diameter of 4 mm. However, it turned out that major disadvantages of the instrument are as follows: OCT imaging was the best technique available for this purpose when compared with other imaging methodologies. 1. the device can only measure the anterior surface, and A special viewing chamber from Bausch and Lomb (IVC-12) hence, corneal thickness cannot be measured; was required to measure the entire tomography of donor 2. the corneal parameters cannot be measured when corneas using clinical OCT devices. Thanks to their special contained in a preservation medium. construction, the use of these commercially available cham- Portable Placido ring–based topographers can be seen bers meant that the corneal grafts did not require any as a miniaturization and potential improvement over tabletop mechanical fixation during imaging. The viewing chamber

Copyright Ó 2016 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. Janunts et al Cornea Volume 35, Number 5, May 2016

FIGURE 6. Tomographic analysis of a keratoconus cornea excised from a keratoconus patient undergoing a penetrating keratoplasty. A 3- dimensional OCT scan was performed using a CASIA corneal tomographer (Tomey) (A). A pachymetry distribution map (B) was generated by calculating a thickness profile for each cross-sectional OCT image. has a flat, optically transparent, front plate facilitating In fact, individual microscopic observation, which is alignment of the cuvette perpendicular to the optical axis. currently used in eye banks, is insufficient for such a chal- Overall, the OCT measurement of donor corneas in lenging diagnostic procedure. The donor corneal screening cuvettes was very straightforward and uncomplicated. Inter- procedure presented here is meaningful and applicable before preting anterior corneas as spherical surfaces rather than all types of keratoplasty. This is of particular relevance for aspherical surfaces was found to be more reliable. This is femtosecond laser–assisted keratoplasty, where an exact because the aspherical fit algorithm did not incorporate the measurement of the midperipheral thickness is of utmost rotation angle (astigmatism), but rather the mathematical importance for planning cutting profiles (eg, mushroom, top fitting algorithm instead interpreted the surface as a rotation- hat, zigzag). In clinical practice, the thickness is estimated ally symmetric aspheric surface for reasons of simplicity. using ultrasound to manually measure the graft at specific However, the samples were completely realigned for the points: 1 central and 4 peripheral spots. Thickness distribution reproducibility measurements. We assume that the change maps of the entire cornea will provide accurate calculation in rotation angle during each measurement could have strategies for planning more sophisticated cutting shapes in influenced the aspheric fit parameters. Nevertheless, all femtosecond laser–assisted keratoplasty. Donor corneal thick- other acquired and extracted data were found to be ness maps may essentially contribute to further optimization highly reproducible. of the trephination strategy for each individual donor cornea A set of various screening methods have been proposed in excimer laser–assisted keratoplasty.13,14 Moreover, the for the correct tomographic analysis of corneal grafts before potential benefits of geometrically matching donor and patient keratoplasty. Particularly, by an adaptive edge-enhanced corneas may be elimination of anterior and posterior step method, any kind of scars, accompanied by morphological formation, exclusion of an interface (gaping) at the graft–host changes within the cornea, may be visualized automatically. junction, and reduced leakage. The technique allows highlighting fine details within the Further investigation is needed to define the swelling corneal tissue and any signs of previous surgeries, which factor for corneas in preservation medium. Having obtained cannot be perceived by the human eye when inspecting cross- a coefficient for the swelling behavior, and using the sectional raw images in most cases. Moreover, the thickness screening method presented herein, we could estimate the distribution map was capable of indicating thin corneas, and postoperative thickness map from measurements made when those with extreme topographic irregularities. the cornea is kept sterile in the preservation medium.

Copyright Ó 2016 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. Cornea Volume 35, Number 5, May 2016 Donor Corneal Tomography Using Anterior Segment OCT

FIGURE 7. Tomographic analysis of a donor cornea with an abnormal thickness distribution. A 3- dimensional OCT scan was performed using a CASIA corneal tomographer (Tomey) (A). A pachymetry distribution map (B) was generated by calculating a thickness proﬁle for each cross-sectional OCT image.

The method presented in this work is suitable for 5. Stoiber J, Ruckhofer J, Hitzl W, et al. Evaluation of donor tissue with detecting and discounting donor corneas that are unsuitable a new videokeratoscope: the Keratron Scout. Cornea. 2001;20:859–863. 6. Priglinger SG, Neubauer AS, May CA, et al. Optical coherence for future transplants, such as those with keratoconus or the tomography for the detection of laser in situ keratomileusis in donor presence of damage from previous refractive surgery corneas. Cornea. 2003;22:46–50. or keratoplasty. 7. Wolf AH, Neubauer AS, Priglinger SG, et al. Detection of laser in situ keratomileusis in a postmortem eye using optical coherence tomography. J Cataract Refract Surg. 2004;30:491–495. 8. Lin RC. Optical Coherence Tomography for the Screening of Donor ACKNOWLEDGMENTS Corneas and Examination of the Retinal Nerve Fiber Directional We acknowledge the financial support provided by the Reflectance [dissertation]. Case Western Reserve University; 2006, p 46. Dr. Rolf M. Schwiete Foundation. 9. Lin RC, Li Y, Tang M, et al. Screening for previous refractive surgery in eye bank corneas by using optical coherence tomography. Cornea. 2007; 26:594–599. REFERENCES 10. Möller T, Machiraju R, Mueller K, et al. Evaluation and design of filters 1. Michaeli-Cohen A, Lambert AC, Coloma F, et al. Two cases of using a Taylor series expansion. IEEE Trans Vis Comput Graph. 1997;3: a penetrating keratoplasty with tissue from a donor who had undergone 184–199. LASIK surgery. Cornea. 2002;21:111–113. 11. Tavakol M, Dennick R. Making sense of Cronbach’s alpha. Int J Med 2. Ehrenhaus MP, Schanzlin DJ, Weissman L, et al. Eye banking error: Educ. 2011;2:53–55. report of a donor cornea with a soft contact lens left in place. Cornea. 12. Leung WC. Statistics and Evidence-Based Medicine For Examinations. 2006;25:359–360. LibraPharm Limited; 2001, p 84. 3. Kang SJ, Schmack I, Edelhauser HF, et al. Donor corneas mis- 13. Seitz B, Langenbucher A, Naumann GOH. Cornea and external eye identified with prior laser in situ keratomileusis. Cornea. 2010;29: disease. In: Larkin T, Reinhard DFP, eds. Essentials of Ophthalmology. 670–673. Springer; 2006:135–136. 4. Ousley PJ, Terry MA. Use of a portable topography machine for 14. Ambrósio R Jr, Alonso RS, Luz A, et al. Corneal-thickness spatial profile screening donor tissue for prior refractive surgery. Cornea. 2002;21: and corneal-volume distribution: tomographic indices to detect kerato- 745–750. conus. J Cataract Refract Surg. 2006;32:1851–1859.