2.1.

Corneal Imaging with Gabor-Domain Optical Coherence Microscopy

GD-OCM (see Fig. 1) is based on spectral-domain OCT, with the particularity of providing a volumetric cellular resolution in the cornea as well as an extended depth of imaging beyond the depth of focus of the objective, necessary to image the full depth of the cornea at cellular resolution. For this study, the light source was a superluminescent diode laser centered at 840 nm with 100 nm FWHM (BroadLighter D-840-HP-I, Superlum, Ireland), which enabled a theoretical axial point spread function (PSF) of $2.2\mu \mathrm{m}$ in the cornea (average refractive index of 1.387 at 840 nm). The interferometer consisted of a $2\times 2$ fiber optics coupler (FC850-40-50-APC, Thorlabs) with a 50:50 split ratio. The two outputs of the interferometer are spliced to two custom fiber collimators (LPC-04-840-5/125 with a 3-m-long and 3-mm OD cable, Oz Optics, Canada) to serve as reference and sample arms. The reference arm was equipped with a polarization controller (FPC030, Thorlabs), and a dispersion compensation block (a diffractive grating—GR25-0608, a focusing lens—AC254-050-B, and a flat mirror—PFSQ10-03-G01, Thorlabs) to compensate the dispersion from fiber mismatch between the reference and sample arms, as well as the dispersion from the GD-OCM objective.³⁸ The custom spectrometer was based on the Czerny–Turner model³⁹ and was made of an achromatic doublet—L1 (AC127-019-B-ML, Thorlabs), two concave mirrors—CM1 and CM2 (CM508-200-P01 and CM508-150-P01, Thorlabs), a diffractive grating—DG1 (GR25-1208), a cylindrical lens (LJ1105L2-B, Thorlabs), and a high-speed CMOS line camera (spl4096-140 km, Basler Inc., Exton, Pennsylvania) operating at the line rate of 70 kHz.

Fig. 1

Schematic representation of GD-OCM. CM, concave mirror; M, flat mirror; PC, polarization controller; FC, fiber collimator; DG, diffractive grating; L, lens; and CL, cylindrical lens.

The microscope objective probe incorporated an electrowetting liquid lens, which used dynamic focusing to image different depths of the cornea while maintaining a lateral resolution of $\sim 2\mu \mathrm{m}$ within the imaging depth of up to $\sim 2\mathrm{mm}$.³³ For this lateral resolution, the depth of focus of the scanning lens was $\sim 100\mu \mathrm{m}$ at any set focus. Consequently, to image all the layers of the cornea at each location, at the full lateral resolution capability of the system, a $1\mathrm{mm}\times 1\mathrm{mm}$ field of view, corresponding to $1000\times 1000$ A-scans, was scanned multiple times with different focal lengths of the liquid lens. Only the desired zone (the confocal parameter of the microscope) in each acquired volume contributed to the final 3-D image.^34,35 For high speed imaging, the acquisition of each zone was done simultaneously with the processing and fusing of the precedent zones, using a parallelized multigraphics processing unit architecture.⁴⁰ A negligible 20 ms delay was allowed between the acquisition of the different zones, corresponding to the automatic adjustment and stability of the liquid lens. The total time for data acquisition was thus equal to the number of zones acquired, multiplied by the acquisition time of each zone (15 s).⁴⁰ In this study, the average number of zones required to image the full depth of pathologic corneas was eight, yielding a total acquisition time of 2 min. It is worth noting that for a targeted layer of the cornea, a single volume with a unique set focus near the layer of interest could be sufficient, yielding 15 s for the acquisition of a $100\text{-}\mu \mathrm{m}$-depth-resolved volumetric image of the depth of interest. During the GD-OCM imaging, the optical contact between the microscope objective and the cornea ($\sim 100\text{-}\mu \mathrm{m}$ gap) was made using saline. The incident power of the cornea was measured to be 2.2 mW, which is at least three times below the threshold for thermal tissue damage and in vivo eye safety by American National Standards Institute (ANSI).⁴¹ Furthermore, the beam focusing on the cornea is widely spread out when it reaches the retina, further reducing any risk of retinal phototoxicity. The 3-D images of a $1\mathrm{mm}\times 1\mathrm{mm}$ ($1000\times 1000$ A-scans) field of view were captured at the center of the corneas using a custom Python program on Linux platform. The free open-source image-processing platform ImageJ⁴² was used for analyzing all data and for optimizing the contrast of the images. The contrast enhancement was performed using a volumetric histogram equalization that yielded 3-D images with a full dynamic range of 16 bits. In addition, Voxx,⁴³ the volume rendering program for 3-D microscopy developed at the Indiana Center for Biological Microscopy (Indianapolis, Indiana), was used to visualize the datasets as solid 3-D images and as two-dimensional (2-D) images of the different layers.

2.2.

Corneal Endothelium Flattening and Analysis

As an OCT-based system, GD-OCM not only generated 2-D cross-section images to evaluate the thickness of the layers but also produced 2-D en face (cut-through) images of each layer, which were essential to evaluate the cellular structures of the layers across a large field of view. As mentioned in the introduction, the curvature of the cornea often compromises the visualization of the entire layer on a single 2-D en face image. In addition, the pathological conditions, such as FECD and KC, induce a significant folding of the endothelium in vivo, which makes the visualization of ECs even more challenging. This constraint is more pronounced on excised human corneas, as the endothelium becomes considerably folded ex vivo due to the loss of intraocular pressure as well as stromal swelling. In this study, we have developed an algorithm to flatten the entire 3-D images of the cornea relative to the endothelial layer. The first step of the algorithm consists of layer segmentation and computation of the 3-D surface of the endothelium. Several approaches for corneal layer segmentation have successfully been developed in the context of the thickness evaluation of corneal layers.⁴⁴45.^–46 However, to the best of our knowledge, these implementations have often been limited to 2-D cross-sectional images and not extended to en face layer visualization. In the proposed framework, the layer of the endothelium was detected for each B-scan of the 3-D volume by applying a modified version of a peak detection algorithm⁴⁷ to individual A-scans. Figure 2(b) presents the depth profile of the A-scan along the blue line in Fig. 2(a), in which the endothelium is marked with the red star. The algorithm was extended to all the frames of the 3-D image to generate the 3-D surface of the endothelium. A polynomial fit was applied to the resulting peaks of the endothelium to smooth the surface and eliminate eventual artifacts from the peak detection. The red curve in Fig. 2(a) shows the overlap between the endothelium and the resulting polynomial fit on a given frame. The second step of the algorithm consisted of pixel shifting. For each A-scan of the volume, the difference in terms of number of pixel between the corresponding position of the endothelium and the apex of the overall 3-D surface from the polynomial fit was evaluated. The entire A-scan was then circularly shifted by the difference in the pixel number to reconstruct the final 3-D image with a flat endothelium. Figure 2(c) shows the result of the pixel shifting on one frame, in which the endothelium now appears flat. Figures 2(a)–2(i) present the case of a cornea, in which the initial endothelial surface was folded [Figs. 2(a), 2(d), and 2(g)]. The endothelial surface was computed from the initial volume [Figs. 2(e) and 2(h)]. Unlike the initial cut-through en face image [Fig. 2(d)] that includes both the endothelium and the DM, all ECs can be visible on a single 2-D en face image of the endothelium after the flattening [Fig. 2(f)]. Figures 2(j)–2(l) illustrate the case of a FECD cornea, for which the flattening algorithm helped to visualize the $1\mathrm{mm}\times 1\mathrm{mm}$ en face view of ECs.

Fig. 2

Flattening of the 3-D image of the endothelial layer in the cornea to facilitate further endothelial analysis using standard segmentation methods and pixel shifting. (a) The endothelium layer mapping on a given frame. EP, epithelium; BM, Bowman’s membrane; ST, stroma; DM, Descemet’s membrane; EN, endothelium; and EC, endothelial cell. (b) The depth profile of a given A-scan, which illustrates the identification of the endothelium layer using the peak detection algorithm. (c) Result of the flattening algorithm applied to the frame in (a), on which the endothelium is flat. (d) The initial endothelial layer. The cut-through image included the endothelium and the DM due to the folding of the endothelium. (e) The corresponding 2-D surface map of the endothelium. (f) The endothelial layer after flattening. Unlike the initial image (d), the cut-through of the endothelium after flattening (f) did not include the DM. (g), (h), and (i) The 3-D images of the initial volume, the endothelium surface map, and the resulting volume after flattening, respectively. (j), (k), and (l) The case of a FECD corneal endothelium before (j) and after (l) the flattening. (k) The mapping of the endothelium in this case.

2.3.

Study Samples

One control cornea and eight corneas with four well-characterized diseases were analyzed in this study: FECD, PBK, KC, and type IIIA LCD. FECD is the second most prevalent and well-established corneal dystrophy that affects 4% of the American population over 40 years of age. This condition is associated with an unexplained appearance of drops (guttae, corresponding to excrescences of collagen fibers) at the posterior surface of the DM, yielding the increase in DM thickness, the decrease in ECD, and the loss of the endothelium pump function leading to corneal edema. PBK is a decompensation of the endothelium following complication of cataract surgery, which can also lead to corneal edema and epithelial bullae. KC, the most common corneal dystrophy worldwide that mainly affects teenagers and young adults, is characterized by a progressive thinning of the cornea that affects the corneal shape, with subtle structural changes from a microscopic point of view. Type IIIA LCD is a genetic stromal corneal dystrophy that starts as an accumulation of amyloid deposits, which then branches as linear opacities in and under the Bowman’s membrane and slowly spreads deeper into the stroma. Table 2 presents the characteristics of the corneas investigated.

Table 2

Characteristics of corneas observed with GD-OCM. The stage of FECD has been evaluated according to the Krachmer grading scale.48

All procedures to obtain the corneas conformed to the tenets of the Declaration of Helsinki for biomedical research involving human subjects. Central corneal buttons of patients were removed at the time of treatment with penetrating keratoplasty and were collected as per the usual protocol in place at the University Hospital of Saint Etienne, France. This protocol was written by the hospital’s commission for clinical research and innovation and accepted by the local ethics committee (CPP Sud Est I, CHU Saint Etienne, France, IRB00010220). Normal corneas assigned to scientific use were procured from bodies donated to science (Laboratory of Anatomy, Faculty of Medicine, Jean Monnet University, Saint Etienne, France), as permitted by French law. Each donor volunteered their body and gave written consent to the Laboratory of Anatomy; no additional approval by the ethics committee was required. The disease stage of corneas with FECD and PBK was clinically evaluated using SM, and classified based on the Krachmer grading scale as well as the confluence of guttae. Patients with LCD and KC were evaluated using the clinical tests light microscopy and corneal tomography, respectively. Finally, a normal cornea from a donor was also evaluated using light microscopy after dissection. All corneas were fixed in 4% paraformaldehyde (Sigma, Saint-Quentin Fallavier, France) for 12 h, rinsed in balanced salt solution (Alcon, Rueil Malmaison, France), and shipped to Rochester, New York for experimentations using GD-OCM. Except for the normal cornea chosen as the control, which was fixed 9 h postmortem, the fixation of all corneas was performed within 1 h after retrieval.

3.1.

Gabor-Domain Optical Coherence Microscopy Performance Evaluation

After system optimization for dispersion and polarization [Fig. 3(a)], the axial resolution of GD-OCM was evaluated using the glass window of the probe as the sample. The mirror of the reference arm was then axially moved to capture the PSF at different depth locations. The average axial PSF of $4\mu \mathrm{m}$ in air (equivalent to $2.8\mu \mathrm{m}$ in the cornea) was measured across the 1.2-mm depth. The sensitivity roll-off was about two times lower at the depth of 1.2 mm compared to 0.2 mm, which corresponds to about 3-dB sensitivity drop [Fig. 3(b)]. Figure 3(c) highlights the PSF at the depth of about 0.3 mm. The lateral resolution of the system was also evaluated by imaging a standard USAF 1951 target (R1L1S1P, Thorlabs) at the same depth of 0.3 mm. Figure 3(d) presents the $1\mathrm{mm}\times 1\mathrm{mm}$ en face view of the target imaged with GD-OCM, showing that the system can clearly resolve features of group-7 and element-6, which correspond to a resolution of $2.2\mu \mathrm{m}$.

Fig. 3

Resolution and sensitivity of GD-OCM. (a) Interference pattern at the depth of $\sim 0.3\mathrm{mm}$. (b) Sensitivity roll-off in the linear scale showing an almost linear drop of the sensitivity with the depth. (c) PSF at the depth of $\sim 0.3\mathrm{mm}$ with a FWHM of $\sim 4\mu \mathrm{m}$ in air ($2.8\mu \mathrm{m}$ in the cornea). (d) En-face view of a standard USAF 1951 target imaged with GD-OCM. A zoom on the smallest group-7 and element-6 shows that the system can clearly resolve the features of this group, which corresponds to a lateral resolution of $2.2\mu \mathrm{m}$.

All the corneas were then successfully imaged with GD-OCM. The average thickness of the different layers was computed to evaluate the relative change in thickness at different stages of the disease. Except for the display of the DM and endothelial layer that leveraged the flattening algorithm, all the images are presented in their initial shape that included corneal curvature. As can be seen in Fig. 2(a), the folding of the posterior layers due to FECD, for instance, does not translate to the anterior cornea, and as such, compensating the anterior layers from the endothelium folding will introduce undesirable folding in the anterior layers [Fig. 2(c)]. A similar approach can be used to compensate the curvature or eventual folding of the anterior layers using the epithelial–Bowman’s membrane interface as the target for the flattening algorithm in case one is interested in visualizing the full en face view of corneal epithelial nerve plexus, for instance.^49,50

3.2.

Microstructural Organization of the Control Cornea

A control cornea from a 61-year-old donor was first evaluated using GD-OCM that provided the high-resolution 3-D image of a 1-mm² central cornea, as well as high contrast cross-sectional and en face images of the different layers of the cornea (see Fig. 4). The anterior [Fig. 4(e)], middle [Fig. 4(f)], and posterior [Fig. 4(g)] stroma represent $\sim 10\%$, 50%, and 90% of the stromal thickness, respectively. These images demonstrate the capacities of GD-OCM to assess the microstructural organization of the different layers of the cornea.⁵¹ We believe that the nonhexagonal pavement of ECs [Fig. 4(h)] may have been caused by the fixation. Indeed, Fig. 4(i) presents the case of a fresh (unfixed) ex vivo human cornea imaged with GD-OCM, which displays the hexagonal shape of the ECs. For this comparative study, it was necessary to use a fixed cornea as baseline.

Fig. 4

Images of the control cornea using GD-OCM. (a) A 3-D high-resolution image of the central cornea. (b) Full-depth cross-sectional image of the cornea at its physiological state. The average thickness of the different layers of the cornea is computed: EP, epithelium; BM, Bowman’s membrane; ST, stroma; DM, Descemet’s membrane; and EN, endothelium. (c) A 2-D view of the epithelium. (d) A cut-through image revealing simultaneously the Bowman’s membrane, basal epithelial cells, and the anterior stroma, given the corneal curvature. En face images of the (e) anterior, (f) middle, and (g) posterior stroma reveal stromal keratocytes (white arrows). (h) A 2-D view of ECs over the full field of view after the flattening of the endothelium. The hexagonal shape of ECs seems to be affected by the fixation. (i) Comparative 2-D view of ECs of a fresh human cornea displaying the hexagonal shape of ECs.

3.3.

Evaluation of Corneas with Fuchs Endothelial Corneal Dystrophy

The disease stage of all corneas was determined clinically. Grades 1 and 2 FECD were not reported in the present study, as these grades were not indications for penetrating keratoplasty, and such cases did not occur in the donor corneas retrieved during this period.

The first case of FECD evaluated in the study was an excised cornea procured from a 92-year-old male donor. The cornea was initially classified as normal during the routine evaluation prior to GD-OCM imaging, and the donor had no previous history of eye disease. First, a $1\text{-}{\mathrm{mm}}^2$ area of the central cornea was imaged using GD-OCM (see Fig. 5). Except for the change in reflectivity and size of the keratocytes located in the posterior stroma [Fig. 5(f)], the rest of the layers appeared normal, including the endothelial–Descemet complex. The elongated shape and hyperactivation of keratocytes in the posterior stroma suggested some abnormalities³⁵ that triggered the need for further investigations. After acquisition of multiple areas, guttae were found only in one localized paracentral area [Fig. 5(h)], suggesting a minor form of FECD.

Fig. 5

Images of a minor form of FECD cornea acquired using GD-OCM. (a) Full-depth cross-sectional image of the cornea with normal physiological thickness. (b) A 2-D en face image of the epithelium. (c) The transition of Bowman’s layer reveals epithelial cells (red arrows), the Bowman’s membrane, and the anterior stroma. Stromal keratocytes (white arrows) are identified in the (d) anterior, (e) middle, and (f) posterior stroma. Abnormally larger keratocytes are observed in the posterior stroma (f). No guttae is found on the central corneal endothelium (g). (h) A 2-D view of the endothelium in the only paracentral area where a few guttae are located (white arrowheads). The two black holes on (b), (c), and (d) are believed to be artifacts from bubbles on the window of the microscope, which create shadows that seem to disappear as we go deeper into the stroma.

For the second case, a corneal button with FECD at grade 3 (moderate stage) was excised from a patient (69 years old) during penetrating keratoplasty for FECD and was imaged using GD-OCM (see Fig. 6). The cornea was thicker than the normal cornea, with specifically significant increase in the stroma and DM thicknesses [Fig. 6(a)]. The hyperactivation and elongation of keratocytes appeared to be more pronounced in the middle stroma [Fig. 6(d)], compared to the posterior stroma [Fig. 6(e)]. In addition, the elongation of keratocytes in this case was less pronounced in the posterior stroma, as compared to the early stage of FECD (see Fig. 5). Several guttae were present in the central corneal endothelial–Descemet complex [Figs. 6(f) and 6(g)]. We observed that bright spots at the endothelial surface [Fig. 6(g)] corresponded to guttae covered by ECs. These bright spots were detected in the GD-OCM images due to the high gradient of refractive index between the extremely thin layer of the EC cytoplasm located at the top of the guttae and the guttae themselves. In contrast, a cut-through view of the guttae showed hyporeflective rounded formations surrounded by hyper-reflective bands, known to be cellular bodies of ECs.

Fig. 6

Images of a moderate stage of FECD cornea acquired using GD-OCM. (a) Full-depth cross-sectional image of the cornea with increased thicknesses of stroma and DM. DM is about twice as thick ($\sim 21.3\mu \mathrm{m}$) compared to the control ($\sim 11.5\mu \mathrm{m}$). (b) The transition of Bowman’s layer showing a sub-basal epithelial corneal nerve. Stromal keratocytes (white arrows) are visible in the 2-D en face images of the (c) anterior, (d) middle, and (e) posterior stroma. (f) A cut-through of the transition between the DM and the endothelium from the reconstructed volume with a flat endothelium, which highlights an area filled with densely packed guttae. (g) A 2-D view of the endothelium showing that guttae at this stage are covered by ECs as can be demonstrated by comparing (f) and (g) (white arrowheads).

The third case of FECD was a corneal button at an advanced stage (see Fig. 7). Stroma and DM thickening was also observed in this case. The elongated shape and hyperactivation of keratocytes appeared to have moved further toward the anterior stroma and was more pronounced in the anterior side of the stroma [Figs. 7(c), 7(d)], compared to the middle [Fig. 7(e)] and posterior [Fig. 7(f)] stroma. Guttae were highly confluent across the $1\text{-}{\mathrm{mm}}^2$ central cornea [Fig. 7(g)] and were also easily observed from the 2-D view of the endothelium [Fig. 7(h)].

Fig. 7

Images of an advanced stage of FECD cornea acquired with GD-OCM. (a) Full-depth cross-sectional image of the cornea with increased thicknesses of stroma and DM. Compared to the control, DM is more than two times thicker ($\sim 26.1\mu \mathrm{m}$). (b) A 2-D en face image of the transition of Bowman’s layer presenting the epithelium, the Bowman’s membrane, and the anterior stroma. Keratocytes (white arrows) are present in the (c) anterior stroma, (d) 30% of the stroma thickness where keratocytes are the most active, (e) the middle, and (f) the posterior stroma. Dark lines in (e) and (f) represent the artifacts associated with the folding of the stroma in the storage solution. (g) A cut-through of the transition between the DM and the endothelium from the reconstructed volume with a flat endothelium, where guttae (white arrowheads) are more confluent and easily observed on the whole surface. The reflectivity of the endothelial surface (h) indicates that some guttae remained covered by EC cytoplasm.

Finally, a corneal button with FECD at the latest stage was evaluated using GD-OCM (see Fig. 8). Small epithelial bullae as a consequence of advanced edema reaching the epithelium were observed [Fig. 8(b)]. The increased thicknesses of the stroma and DM were also confirmed in this case. An apparent loss in the number of keratocytes was observed in the whole stroma and specifically in the posterior stroma, where keratocytes were almost absent [Fig. 8(f)]. Large guttae were visible on the whole endothelial surface with almost no remaining ECs [Fig. 8(g)].

Fig. 8

Images of a cornea with the latest stage of FECD acquired with GD-OCM. (a) Full-depth cross-sectional image of the cornea with increased thicknesses of stroma and DM. DM is more than two times thicker ($\sim 24.8\mu \mathrm{m}$) than the control. An epithelium detachment that occurred during cornea handling can be observed on (a). (b) A 2-D view of the epithelium presenting the development of bullae. The 2-D en face images of the (c) anterior stroma, (d) 30% of the stroma thickness, (e) the middle, and (f) the posterior stroma present a few keratocytes (white arrows). (g) A 2-D view of the endothelium with numerous large guttae (fusion of several guttae, white arrowheads) and absence of ECs, which presents artifacts due to the intensity saturation or interference fringes.

3.4.

Evaluation of Corneas with Pseudophakic Bullous Keratopathy

To investigate cases of corneal edema not caused by FECD, we evaluated two cases of PBK (see Fig. 9). In both cases, we observed a high increase in corneal thickness. The elongated shape and hyperactivation of keratocytes appeared to be localized in the posterior stroma [Figs. 9(f) and 9(m)], whereas the anterior [Figs. 9(d) and 9(k)], and middle [Figs. 9(e) and 9(l)] stromal keratocytes appeared normal. In the first case, it was difficult to establish a clear boundary between ECs and DM, which appeared to be thinner than the control and presented filamentous-like structures [Fig. 9(g)], presumably similar to a case of a non-guttate endothelial degeneration, as previously described by Abbott et al.⁵² and Yuen et al.⁵³ No evidence of guttae was found in this case [Fig. 9(g)]. The second case was an unusual situation, likely an early stage of FECD that was decompensated after cataract surgery. This case revealed an advanced stage of corneal edema with the development of epithelial bullae [Fig. 9(i)]. Guttae-like structures were observed in the endothelium [Fig. 9(n)]. However, no increase in DM thickness, a feature associated with FECD, was observed in these two cases.

Fig. 9

GD-OCM images of two cases of PBK without the thickening of DM. The first case of endothelial dystrophy (a–g) presented no evidence of guttae in the endothelial–Descemet complex (g). (b) A 2-D en face image of the epithelium highlighting the epithelial cells and no evidence of bullae. (c) The transition of Bowman’s layer. Stromal keratocytes (white arrows) are visible in the (d) anterior, (e) middle, and (f) posterior stroma. (h)–(n) The second case of endothelial dystrophy reveals the development of (i) epithelial bullae and (n) guttae-like structures in the endothelial–Descemet complex. (j) The transition of Bowman’s layer. Keratocytes (white arrows) are also present in the (k) anterior, (l) middle, and (m) posterior stroma.

3.5.

Evaluation of a Keratoconus Cornea

GD-OCM image of a KC cornea revealed a significant amount of stromal striae⁵⁴ (see Fig. 10) and other artifacts caused by the low rigidity of the thin corneal button, which easily deformed and retracted in a Petri dish. Keratocytes were less visible than in the control cornea. ECs were morphologically normal.

Fig. 10

Images of a KC cornea. (a) Bright-field microscopic image of the endothelium highlighting EC nuclei staining. (b) Full-depth cross-sectional image of the cornea acquired using GD-OCM. The cornea is thinner than the normal physiological state, with a particularly thin Bowman’s layer and folding of the endothelium (white arrowheads), presumably some stromal striae. (c) The transition of Bowman’s layer. The middle (d) and posterior (e) stroma present significant folding artifacts with reduced number of keratocytes. (f) A 2-D view of the endothelium after numerical flattening, which highlights the normal morphology of ECs. Lower-level folding still remains after the flattening due to the limitation of the polynomial fitting to approximate the curvature of sharp folds on the surface.

3.6.

Evaluation of a Cornea with Lattice Cornea Dystrophy

A corneal button from a patient with LCD was evaluated using GD-OCM (see Fig. 11). Images revealed lattice branching filaments in the anterior stroma. These filaments appeared as reflective lines creating shadows in the posterior cornea. As compared to stromal striae that are localized in depth, the shadow in this case is starting from the location of the lattice fibers in the anterior stroma and going all the way down to the endothelium. We can also see that the shadow maintains the same shape of the lattice fibers as we go deeper into the stroma, which is not typical of stromal striae.

Fig. 11

Images of a type III LCD. (a) Light macroscopy image of the LCD cornea presenting opaque lattice branching $\sim 100\text{-}\mu \mathrm{m}$ wide. (b) Full-depth cross-sectional GD-OCM image of the cornea with vertical shadows of lattice branches (white arrowheads). (c) The transition of Bowman’s layer presenting some disruption and whitish deposits in the basal epithelium, likely scars after the healing of epithelial ulcers, characteristic of LCD (white stars). (d) The anterior stroma reveals a few lattice branching filaments (white arrowheads), which are amyloid deposits or abnormal protein fibers. White arrows indicate stromal keratocytes. (e) The middle stroma presents the shadows of lattice branches. (f) The endothelium that is not processed with the flattening algorithm due to lattice fiber shadows. ECs, nonetheless, seem to have preserved the normal morphology.