Imaging of Birefringent Properties of Keratoconus Corneas by Polarization-Sensitive Optical Coherence Tomography

Abstract

purpose. To investigate and map the polarizing properties of keratoconus corneas in vitro and to compare the results with those obtained in normal corneas.

methods. Corneal buttons of five keratoconus corneas were investigated by polarization-sensitive optical coherence tomography (PS-OCT). The instrument measures backscattered intensity (conventional OCT), retardation, and (cumulative) slow axis distribution simultaneously. Three-dimensional (3-D) data sets of the polarizing parameters are recorded, and two-dimensional (2-D) cross-sectional images as well as en face images of the distribution of these parameters at the posterior corneal surface are derived. The results are compared to similar maps obtained from normal corneas.

results. Compared with normal corneas, the retardation and slow axis orientation patterns are heavily distorted in keratoconus corneas. Larger areas of increased and decreased retardation can be found in keratoconus corneas, markedly increased retardation (up to >50°) can especially be found near the rim of corneal thinning. Contrary to normal corneas, regions where the slow axis markedly changes with depth (by up to 50°–90°) are observed in keratoconus.

conclusions. The observed changes in the cornea’s birefringence properties indicate a change in the arrangement of collagen fibrils in the corneal stroma associated with keratoconus. PS-OCT may be a useful tool for the study and diagnosis of corneal disease.

The tissue of the human cornea is arranged in several layers. More than 90% of its thickness is made up by the stroma, which consists of approximately 200 lamellae with a thickness of 1.5 to 2.5 μm. Each of these lamellae is composed of parallel collagen fibrils embedded in an optically homogenous ground substance. ^{1 2} The fibrils in successive lamellae are usually oriented at large, approximately orthogonal angles to each other. X-ray diffraction studies have revealed two preferred directions of fibril orientation in the cornea. Approximately 60% of the fibrils are oriented within the 45° sectors around the inferior–superior and nasal–temporal directions, whereas ∼40% are oriented in the oblique areas between those sectors. ^{3 4} The regular arrangement of fibrils within each lamella is responsible for the transparency of the tissue, ^{1 5} whereas the orientation of successive fibril layers is important for establishing the mechanical stability of the cornea. ^{6 7}  

Because of the critical dependence of the corneal shape and mechanical stability on the orientation of fibrils in corneal lamellae, diseases like keratoconus, in which corneal shape changes, are associated with changes in the regular fibril arrangement. ^{4 8} Keratoconus is a common dystrophy of the cornea and is characterized by a progressive thinning and ectasia of the central cornea that causes myopia and irregular astigmatism. With an incidence of approximately 50 to 230 per 100,000 of the population, ^{9 10} it is a frequent indication for keratoplasty. 

Although the knowledge of details of local fibril orientation is based on electron microscopy studies, ^{1 11} information on structural characteristics of the entire cornea is essentially based on x-ray diffraction methods. ^{3 4 8 12 13} With this technique, an intense beam of x-rays transmits the cornea and is scattered to produce a diffraction pattern characteristic of the ordered structure of the tissue’s constituents. Details on diameter and orientation of the collagen fibrils, integrated over the transmitted volume, can be obtained. With this method, it has been shown that the angle between the preferred collagen orientations in keratoconus corneas were changed from 90° and 180° to 60° and 120°, indicating structural changes in these corneas. ⁴ Furthermore, a recent study using high-intensity synchrotron radiation enabling a transverse resolution of 0.25 mm demonstrated a dramatic change of the gross organization of stromal lamellae and an uneven distribution of collagen fibrillar mass in keratoconus corneas. ⁸ Despite the detailed knowledge on collagen structure that has been provided by x-ray diffraction techniques, this technology has some severe drawbacks: Measurement times are very long (several hours), which, in connection with the ionizing nature of x-rays, prevents the use of this technique for in vivo diagnostic applications. Furthermore, the integral nature of the measurement (being performed in transmission) prevents depth-resolved information. 

Another method that can be used to obtain information on the corneal structure is probing the cornea with optical radiation. Although scattering methods can also be used, ¹⁴ another effect that can be exploited is the birefringence of the cornea. It has long been known that the corneal stroma is optically birefringent, and several papers on corneal birefringence have been published (for an overview, see Ref. ¹⁵ ). Different models on the nature of corneal birefringence have been reported that are partly contradictory (e.g., the cornea was modeled as a uniaxial ¹⁶ and a biaxial crystal). ¹⁷  

In a recent study, we suggested a new model of corneal birefringence. ¹⁸ This model is based on detailed studies of corneal polarization properties with polarization-sensitive optical coherence tomography (PS-OCT), ¹⁹ and on the knowledge of the collagen fibril orientation derived from x-ray diffraction studies. In modeling corneal birefringence, each lamella of the corneal stroma is regarded as a birefringent plate with its slow axis lying along the direction of the collagen fibrils. ²⁰ The fibrils in successive lamellae are usually oriented at large, approximately orthogonal angles to each other, so that the birefringence properties of successive lamellae cancel each other largely. A slight prevalence of one lamella orientation causes a net retardation and a net optic axis orientation. 

This model describes the observation of corneal birefringence with low retardation and a preferred axis near the corneal apex under the assumption that the probing light beam impinges perpendicularly on the corneal surface. Toward the periphery, the beam illuminates the cornea with increasing inclination to the corneal surface’s normal vector (the beam is parallel to the optic axis of the eye). For a nonperpendicular probing beam, the birefringences of successive lamellae no longer compensate each other, and retardation increases toward the periphery. Also the slow axis shows a characteristic pattern with axis orientation varying linearly with azimuth angle. ¹⁸  

PS-OCT is a variant of OCT ²¹ that measures, in addition to intensity, the polarization state of the light backscattered by the sample, allowing three-dimensional (3-D) information to be obtained on various polarizing properties of tissue (e.g., birefringence, ^{22 23 24} diattenuation, ^{25 26} and depolarization, or polarization scrambling). ²⁷ We developed a PS-OCT method that allows the measurement of three parameters simultaneously: reflectivity, retardation, and birefringent axis orientation. ²⁴ In this study, we used our PS-OCT technique to measure and image the polarizing properties of keratoconus corneas in vitro. We obtained 3-D data sets of reflectivity, retardation, and slow axis orientation and derived cross-sectional images of the birefringence parameters, as well as images showing the distribution of these parameters across the posterior surface of the cornea. We compared these results with those previously obtained in normal corneas and, in this article, discuss implications for diagnostic use of PS-OCT. 

Methods

PS-OCT Technology

PS-OCT imaging was performed with a system previously described. ^{18 24} The system is based on a classic bulk optics Michelson interferometer and works in time domain. A superluminescent diode emits a low-coherence beam (center wavelength, 828 nm; full-width-at-half-maximum bandwidth, 22.4 nm; coherence length, 13.5 μm; depth resolution in tissue, ∼10 μm) that is vertically polarized before entering the interferometer. A nonpolarizing beam splitter cube splits the beam into sample and reference beams. The polarization plane of the reference beam is rotated by 45° on double passing a quarter-wave plate, providing equal power to both channels of a polarization-sensitive detection unit. The sample beam transmits a quarter-wave plate oriented to have a circularly polarized beam incident on the sample. After recombination of reference and sample beams, the interfering beams are split by a polarizing beam splitter into the horizontal and vertical detection channel, where they are phase sensitively recorded by separate photodetectors. 

The cornea was illuminated with a power of ∼600 μW. Between the individual A-scans, the sample was moved by a motorized x–y translation stage to obtain A-scans distributed over the cornea along an evenly spaced x–y grid. In this way, 3-D data sets consisting of 80 × 80 A-scans (with a spacing of 75 μm in x and y directions) were recorded, covering a total volume of 6 mm (x) × 6 mm (y) × 2.5 mm (z, optical depth). A-scan rate was 2 Hz. 

To obtain sample reflectivity, retardation δ, and (cumulative) slow axis orientation θ, an algorithm based on a Hilbert transform and on a Jones vector analysis was used. While reflectivity and retardation are obtained from the amplitudes of the two polarization channels, the axis orientation is derived from their phase difference. ^{18 24} On OCT tomograms, the backscattered signal intensity is displayed on a logarithmic false color scale, δ and θ are displayed on linear false color scales. The unambiguous ranges of δ and θ determination are 0° to 90° and 0° to 180°, respectively. 

It should be pointed out that the interpretation of the slow axis data requires some care: (1) The algorithm used to calculate θ causes a 90° change in θ at depth position z, where δ crosses 90° (or multiples of 90°). (2) If several layers with different slow axes are stacked on top of each other, the measured value θ (z), at a given depth position z, is a cumulative, or effective, axis: It is the slow axis of a single layered birefringent plate that would give rise to the same change in polarization state (on double pass of the sample) as the multilayered sample does, down to depth z. To differentiate the axis orientations of individual layers would require that (1) these individual layers be resolved (i.e., are thicker than the coherence length) and (2) an algorithm be used that is based on the information obtained on the overlying layers and propagates the polarization state from layer to layer. 

For comparison purposes, corneal thickness maps were recorded before tissue explantation in the in vivo cornea by the Orbscan system (Bausch & Lomb, Rochester, NY). The Orbscan is based on the principle of slit projection. Two scanning slit lamps project 40 calibrated beams onto the cornea, and the light diffusely backscattered from the anterior and posterior corneal surfaces is measured to obtain topographic maps of the two surfaces. A corneal thickness map is obtained by calculating the difference between these two topographic maps. 

Tissues

For this study, corneal buttons (8 mm diameter) were obtained from five patients after penetrating keratoplasty for keratoconus. The patients were Caucasian, four were male, one was female, and their mean age (±SD) was 48 ± 5 years. The diagnosis of all corneas was advanced keratoconus, all patients were intolerant to contact lenses, and best spectacle corrected visual acuity was <20/40. The surgical procedure was as described before, ²⁸ using an 8-mm Barron-Hessburg trephine to excise the recipient button. Before tissue explantation, Orbscan data were obtained (in one case with only poor quality). Fully informed consent was obtained from all the patients, and the study protocol was in accordance with the Declaration of Helsinki and was approved by the ethics committee of the Medical University of Vienna. 

Before PS-OCT imaging, the corneal buttons were stored in a deswelling medium containing dextran 500 (Medium B). Imaging was performed within a week after explantation (within the time frame used in our previous study on normal corneas ¹⁸ ; a longitudinal monitoring over time confirmed that the polarizing properties of corneal tissue stored in Medium B remain essentially unchanged over 2 weeks). For PS-OCT imaging, the corneas were mounted in a specimen holder consisting of a metal case with a nonbirefringent glass window through which imaging was performed. The case was filled with Medium B to prevent dehydration during imaging. The corneas were measured in an upright position (i.e., such that the vision axis was horizontally aligned with the direction of the sampling beam). The azimuthal rotation of the cornea about the vision axis with respect to its orientation in the intact eye before explantation could not be determined. Some of the corneas were very unstable and deformable and therefore were not suited for measurement in this specimen holder. For these corneas, a modified specimen holder was used in which the corneas were mounted in a horizontal position (i.e., vision axis vertically oriented; the sampling beam was aligned along the vision axis with an additional mirror). To support the tissue, a plastic hemisphere was located below the cornea. To prevent the posterior corneal surface from damage, a thin layer of hyaluronic acid gel (Healon; GE Healthcare, Vienna, Austria) was inserted between the supporting hemisphere and the cornea. After the gel layer and the cornea were inserted, a plastic ring of 6-mm inner diameter was gently laid on the cornea to fixate the rim of the corneal button, thus preventing the tissue from floating. The whole specimen holder was filled with Medium B to prevent the cornea from dehydration during imaging. Two corneas were measured in the upright position, three in the horizontal position. No obvious differences were observed between the two methods. 

For comparison purposes, parts of the results of a previous study ¹⁸ performed on human donor corneas obtained from the eye bank of the General Hospital of Vienna are included in the present study. In that study, full-thickness corneal transplants were used. They were imaged with the same instrument as was used in this study. 

Results

Figure 1shows results obtained from a healthy cornea that were taken for control purposes from a previous study. ¹⁸ Figures 1A and 1Bshow horizontal cross-sectional images of retardation and (cumulative) slow axis orientation, respectively. They were derived from 3-D data sets through the center (apex) of the cornea. The parts of the image shown in gray correspond to regions where the signal intensity was not significantly above the noise level (in this case, a reliable calculation of retardation and axis values was not possible ²⁹ ). The retardation increased in the radial direction. Toward the periphery, the retardation increased with depth (Fig. 1A) . At the margin of the images, a retardation of ∼180° over the full corneal thickness was observed (indicated by a full-color oscillation from blue over red to blue). The cumulative slow axis varied in the radial direction (Fig. 1B) ; however, it was roughly constant over the corneal thickness at a given transverse position (except for the 90° color change at positions where δ > 90°). 

Figures 1C 1D 1Eshow en face PS-OCT images derived from the same 3-D data set. Retardation (Fig. 1C)and slow axis orientation (Fig. 1D)distribution corresponding to the back surface of the cornea are shown. Figure 1Cshows that the retardation was lowest in the center and increased in a radial direction. Figure 1Dindicates that the slow axis orientation varied approximately linearly with the azimuth angle. Because of the 90° color jumps in θ at positions where δ > 90°, the overall image pattern of approximately constant θ at a given azimuth angle was disturbed at the periphery. For a better, undisturbed view of the θ pattern, we derived an additional en face image of θ that corresponds to a position in the middle of the cornea, approximately half way between anterior and posterior corneal surface (Fig. 1E) . Since δ < 90° at this depth throughout the cornea, the axis distribution pattern was undisturbed. A comparison of Figures 1D and 1Efurther indicates that the axis orientation was roughly constant in depth. 

Figure 2shows results obtained from a keratoconus cornea of a 52-year-old male patient (left eye). Figure 2Ais an Orbscan thickness map showing a thinning of the cornea at a slightly decentral position (superotemporal). Figure 2Bshows a thickness map of the corneal button derived from a 3-D OCT data set. The optical thickness was converted to the geometrical thickness by division by the group refractive index n _g = 1.38. ³⁰ A thinning of approximately similar size and shape as in the Orbscan is observed (the azimuthal position of the thinned area deviates due to the unknown azimuthal orientation of the explant). The thickness readings obtained by Orbscan are in general ∼10% larger than those measured by OCT (larger deviations can occur at small localized defects that are below the transversal resolution of Orbscan). The reason may be found in the different measurement techniques or in different hydration states. Since local surface inclination influences the measured birefringence of a healthy cornea, surface elevation data showing the position of the anterior corneal surface were also derived (Fig. 2C) . This surface height plot allows estimation of the local inclination of the cornea and showed a slight indentation of the corneal surface (Fig. 2C , arrow). 

Figure 2Dshows the retardation map derived at the posterior corneal surface. A deviation from the regular pattern of a normal cornea was observed. Figure 2Eshows the slow axis distribution ∼100 μm anterior to the posterior corneal surface (to avoid the color jump in areas where δ had already crossed 90°). The pattern was distorted compared with that in the normal cornea (Figs. 1D 1E) . 

Figures 2F and 2Gshow horizontal and vertical cross-sectional reflectivity images derived from the same 3-D data set (positions indicated by red lines in the en face images). The keratoconus-related thinning of the cornea can be clearly observed, as well as the indentation mentioned earlier. Figures 2H and 2Ishow the corresponding retardation images. A marked asymmetry can be observed: the horizontal cross section (Fig. 2H)shows strong retardation (with exception of the center), whereas the vertical cross section (Fig. 2I)shows only very low retardation (although surface inclination is similar along both sections), with the exception of a small focal area at both sides of the thinning, where an increase in retardation by >50° with respect to the surrounding is observed (arrows). Figures 2J and 2Kshow corresponding slow axis images with deviations from the normal patterns. 

Figure 3shows data obtained from a corneal button of a 43-year-old male patient (left eye). The thickness map (Fig. 3A)shows a marked thinning. The elevation map (Fig. 3B)shows no marked indentation (this cornea was supported by a plastic hemisphere and a Healon gel layer). The en face retardation map obtained from the posterior corneal surface (Fig. 3C)shows a heavily distorted retardation pattern. An interesting observation is a sharp, narrow band of higher retardation approximately surrounding the thinned area observed in Figure 3A . Figure 3Dshows the slow axis distribution at the posterior corneal surface. This pattern is again heavily distorted. 

Figures 3E and 3Fshow horizontal cross-sectional retardation and slow axis images, respectively (position marked by red lines in en face images). The supporting plastic hemisphere below the cornea scrambles the polarization state of the backscattered light. The corneal thinning on the left side of the images is clearly visible. Figure 3Eshows increased retardation near the rim of the thinning (arrows), and low retardation elsewhere (also in steeper regions of the cornea). Figure 3Fshows a sharp axis orientation change at about the middle of the corneal depth (arrows). The color changes from blue-red (blue and red indicate rather similar orientations, the color jump of 180° is caused by 180° wrapping) to green, indicating a change of axis by ∼50° to 90°. Such an axis change with depth is not observed in the normal cornea. 

In the other keratoconus corneas (not shown here) similar observations were made: heavily distorted patterns of retardation and axis orientations, frequently increased retardation near the corneal thinning, and occasional axis orientation changes with depth. 

Discussion

We used PS-OCT to measure retardation and axis orientations of keratoconus corneas in vitro. In a previous paper, ²⁴ we had shown that our instrument provides results with high precision (SD for δ, ∼0.5°, for θ, ∼1.8°) and good accuracy for δ (maximum deviation, ∼3°−4° at δ = 0° and 90°), although there was a systematic deviation of ∼12° for axis orientation, caused by imperfect polarizing elements in the instrument. Because only relative axis orientations are measured in this work (the absolute orientation of the explant with respect to the in vivo eye was not known) we decided not to correct for the systematic offset. Because low signal intensity can degrade the results of PS-OCT measurements, ²⁹ areas with poor signal quality are displayed in gray in our images. An additional factor that degrades image quality in coherent images is speckle noise, which causes the dotted appearance of the PS-OCT images. Averaging techniques can improve the image smoothness, but with the drawback of reduced resolution. Therefore, we decided to present the raw, unsmoothed images. 

The results of our measurements show that the normal cornea has a very characteristic birefringence pattern, ¹⁸ showing low retardation near the corneal apex (at perpendicular beam incidence) that increases approximately radially and symmetrically toward the periphery. At peripheral locations, a pronounced increase of retardation with depth was observed. The slow axis also showed a characteristic pattern: Near the corneal apex, a predominant axis can be observed, whereas off the apex the slow axis varied approximately linearly with the azimuth angle (in case of a sampling beam parallel to the vision axis). There was little change of axis orientation with depth. This pattern is explained by a birefringence model based on a stack of thin birefringent lamellae with two preferential, nearly orthogonal fibril orientations, superimposed on a background of lamellae with generally randomly oriented fibrils that had a slightly preferential orientation. 

The results of this study have shown that this normal pattern is heavily distorted in keratoconus corneas. Typical findings were asymmetric or totally irregular retardation patterns and locally increased retardation near the rim of corneal thinning. The slow axis pattern typically showed larger areas with a predominating range of axis orientations, areas where the axis changed rapidly in a transverse direction, and areas where the axis changed markedly with depth. 

It should be mentioned that the observed corneal birefringence depends on the local inclination of the cornea with respect to the sampling beam. The normal pattern is a result of the cornea’s being composed of individual birefringent lamellae stacked in a deliberate way and of the local inclination of these lamellae. It is not easy to separate these effects. Our goal is to obtain information on how far a change in the structure and/or the arrangement of the lamellae occurs in keratoconus. The change in corneal shape is probably a secondary effect of minor interest for the pathogenesis of keratoconus. Since the corneal explants consist of partly very soft and unstable tissue whose shape can be changed during handling and mounting, we have to take care to compare only those parts of the tissue whose inclination is approximately regular (i.e., we have to avoid areas with indentations or folds; this information is provided by the surface elevation data). Taking this into account, deviations of the keratoconus birefringence patterns from normal subjects are still clearly visible. 

It is, therefore, likely that the observed birefringence changes are caused by a change in the lamellar structure in keratoconus corneas. Such a change was reported almost a decade ago by Daxer and Fratzl, ⁴ who used x-ray scattering methods. Meek et al. ⁸ recently published results obtained by a similar method with improved lateral resolution. They reported a dramatic change in the organization of stromal lamellae and an uneven distribution of collagen fibrillar mass distribution, especially around the apex of the cone. 

Care should be used in direct comparison of x-ray scattering data and birefringence measurements by PS-OCT. Although both x-ray scattering and birefringence depend on the orientation of collagen fibrils, this dependence is different. X-ray scattering shows directly the distribution of collagen fibril orientations within the tissue volume transmitted by the x-ray beam. This is an advantage; however, a drawback is that a depth-resolved measurement is not possible. PS-OCT has the advantage that depth resolution is possible (i.e., changes of the parameters δ and θ with depth can be observed). However, since the individual lamellae cannot be resolved (the depth resolution of our technique is ∼ 10 μm, the lamella thickness ∼ 2 μm), only the birefringence integrated over several lamellae is observed. Because the birefringence orientations of lamellae with orthogonal fibril orientation are also orthogonal, their birefringences cancel each other (at perpendicular beam incidence), and so only an excess of one fibril orientation can be detected, the amount of this excess is proportional to the retardation, and its orientation is equal to the birefringent slow axis. No information is obtained on the orientation of those fibers that cancel each other. This indirect and incomplete information provided by birefringence measurements is certainly a disadvantage of our method (in addition to the sensitivity to local surface inclination). 

Keeping this difference in information provided by both methods in mind, comparisons of the results show quite good agreement. In both cases, heavy distortions of the patterns—scattering and birefringence—are observed in keratoconus corneas. Considerable changes in the local orientation of x-ray scattering patterns correspond to changes in the slow axis orientations observed by PS-OCT. X-ray diffraction demonstrates that lamellar axis orientation tends to curve around the cones. This means a predominant axis orientation in these regions, and a predominant axis orientation leads to increased birefringence which is observed near the rim of thinned areas. In addition, an associated axis orientation change can be partly observed (Figs. 3C 3D) . Finally, keratoconus corneas with locally increased and decreased aligned scattering are observed by the x-ray technique. These areas have a larger and a lower amount of fibrils of a predominant orientation, respectively. In a birefringence pattern, increased and decreased retardation would be expected, respectively, in these areas, and, indeed, such areas can be observed (Fig. 2D) . To summarize, the results of our study support the hypothesis that considerable changes in collagen fibril arrangement are associated with keratoconus. 

Our study has also implications for the use of ophthalmic diagnostic instruments operating with polarized light. In particular, scanning laser polarimetry (SLP) of the retinal nerve fiber layer requires that the corneal birefringence be compensated for. ³¹ The newest generation of SLP has a variable corneal compensator capable of compensating the individual cornea. ³² However, if the instrument is to be used to image through a keratoconus cornea, the corneal birefringence will vary with scan angle, and the corneal compensation will not work properly. Future studies should determine whether this problem also arises in moderate or early-stage keratoconus. 

The PS-OCT instrument used in this study is a slow, time-domain system, therefore it could only be used for in vitro measurements. However, OCT technology has undergone considerable improvement recently. Spectral domain OCT has been shown to have huge sensitivity and speed advantages compared with time domain OCT. ^{33 34} This enables 3-D imaging of the human retina with a data-acquisition time on the order of a few seconds. ^{35 36 37} Recently, we developed a spectral domain PS-OCT system for retinal imaging. ³⁸ This instrument acquires a 3-D PS-OCT data set of the human retina in 3 seconds. The modification of this instrument for the purposes of anterior segment imaging will allow in vivo applications of this technology in cornea diagnostics, as well. Advanced evaluation algorithms quantifying the local birefringence changes compared with the normal pattern and taking into account the information on corneal surface shape and inclination have to be developed to simplify the interpretation of the PS-OCT retardation and axis orientation maps. 

Because PS-OCT is sensitive to changes in the lamellar structure of the cornea, a potential clinical relevance of the technique might be the detection of abnormal corneal biomechanical properties in vivo. These properties can deviate from the normal state as a result of surgery or disease. Until now, there has been no method of determining the biomechanical properties of the cornea in vivo. Potential uses of such measurements might be a better understanding of laser refractive surgery outcomes and the results of nonablative refractive correction such as conductive keratoplasty (CK), and determining other potential factors in ectasia. Furthermore, it may help to understand the impact of corneal physical properties on intraocular pressure measurement and glaucoma risk. 

 Supported by the Austrian Science Fund FWF Grant P 16776-N02.

 Submitted for publication June 29, 2006; revised January 16 and February 13, 2007; accepted May 25, 2007.

 Disclosure: E. Götzinger, None; M. Pircher, None; I. Dejaco-Ruhswurm, None; S. Kaminski, None; C. Skorpik, None; C.K. Hitzenberger, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

PS-OCT images of normal human cornea in vitro. Axis labels: distances (mm). Linear color scales (deg). (A, C) Retardation; (B, D, E) slow axis orientation. (A, B) Horizontal cross sections; (C, D) En face images, posterior corneal surface; (E) en face image, half way between anterior and posterior corneal surfaces. Reprinted, with permission, from Götzinger E, Pircher M, Sticker M, Fercher AF, Hitzenberger CK. Measurement and imaging of birefringent properties of the human cornea with phase-resolved polarization-sensitive optical coherence tomography. J Biomed Opt. 2004;9:94–102.

Figure 2.

 

Images of keratoconus cornea in patient 1 (52 years old). (A) Orbscan thickness map in vivo (color bar: micrometers). (B–K) PS-OCT images in vitro: (B–E) en face images (8 × 8 mm²); (F–K) cross sections (8 mm width × 1.25 mm depth). (B) OCT thickness map (micrometers); (C) anterior surface elevation (micrometers); (D) retardation at the posterior surface; (E) slow axis anterior to the posterior surface. (F) Horizontal and (G) vertical reflectivity images (log scale); (H) horizontal and (I) vertical retardation images; (J) horizontal and (K) vertical slow axis images.

Figure 3.

 

PS-OCT images of keratoconus cornea in vitro, patient 2. (A–D) En face images (6 × 6 mm²); (E, F) cross-sections (6 mm width × 1.05 mm depth). (A) Thickness map (micrometers); (B) anterior surface elevation map (micrometers); (C, E) retardation; (D, F) slow axis.

The authors thank Bernhard Baumann for performing the monitoring measurements on corneal tissue.