Supersonic Shear Wave Elastography for the In Vivo Evaluation of Transepithelial Corneal Collagen Cross-Linking

Abstract

Purpose.: To assess corneal stiffening with supersonic shear wave imaging (SSI) technology in an experimental model of iontophoresis-assisted transepithelial corneal collagen cross-linking (I-CXL).

Methods.: Six rabbits underwent full, central I-CXL in one eye. The contralateral eye served as control. In vivo iontophoresis was used for 10 minutes to perform transepithelial delivery of riboflavin prior to UV-A irradiation. Accelerated UV-A protocol was applied for 9 minutes with a 10-mW/cm² irradiance. Animals were killed and both treated and control corneas were then immediately mounted on a corneal artificial anterior chamber and internal pressure was varied from 15 to 50 mm Hg in 5-mm Hg increments. Swelling was evaluated via central corneal thickness measurements. Ex vivo inflation tests were monitored using SSI technology that provides real-time mapping of the corneal elasticity.

Results.: Corneal yellowing of the central 9-mm diameter area was clearly visible in the iontophoresis area of all treated eyes. Elasticity versus internal pressure revealed significant differences of the change in elasticity coefficient with pressure between I-CXL–treated and control corneas with a mean slope that was 27.1 and 16.9 kPa/mm Hg, respectively (P = 0.029). Differences in elasticity at individual pressure levels between groups were statistically significant above 40 mm Hg (P < 0.05).

Conclusions.: Intraocular pressure variations were the most important limitations for in vivo stiffness monitoring with SSI because stiffness is a function of internal pressure. Supersonic shear wave imaging succeeded in comparing corneas that underwent I-CXL by performing ex vivo inflation tests where pressure was controlled. Iontophoresis-assisted transepithelial corneal collagen cross-linking corneas exhibited increased resistance to pressure rise, indicating stiffening. In vivo I-CXL and ex vivo SSI is an interesting model to evaluate the sole effect of photopolymerization occurring in the CXL process close to physiological conditions.

Introduction

Within the last 10 years, conventional corneal collagen cross-linking (C-CXL) with riboflavin application and UV-A (375 nm) irradiation has become the gold-standard procedure to limit the progression of ectasia in keratoconus patients. ^1,2 From the clinician point of view, C-CXL remains invasive as the corneal epithelium must be removed to allow the photosensitizer to soak the anterior part of the cornea. Moreover, further cases of severe complications have been published corresponding to excessive UV-A effects or to unpredictable healing troubles. ³ Thus, the motivation to avoid epithelium removal during C-CXL is very strong. 

Iontophoresis is defined as a noninvasive technique where a constant electric current is applied to enhance penetration of an ionized substance into a tissue without epithelial removal. ⁴ It has been used in various fields of medicine, ^5–7 for example, local transdermal administration of anesthetics, antiinflammatories, or analgesics, as well as transmucosal administration of antivirals. ⁸ In ophthalmology, the first studies of iontophoresis were performed in the 1940's with the administration of antibiotics for the treatment of bacterial endophthalmitis and keratitis. ⁹ By the end of the century, iontophoresis was extensively investigated for delivering ophthalmic drugs, including dyes, antibacterial, antiviral, antifungal, steroids, antimetabolites, and oligonucleotides. ^10–12 Ocular iontophoresis is still under investigation as an answer to the low bioavailability of drugs after topical administration and to the potential serious complications after intraocular injections used for the treatment of several eye disorders. Recently, the idea of a transepithelial (TE) CXL procedure assisted by iontophoresis (I-CXL) has been considered as a promising method to overcome the lack of corneal epithelial permeability to riboflavin and obtain an adequate intrastromal riboflavin concentration prior to the UV-A irradiation. 

To support the preliminary clinical results and to reach a better understanding of the CXL interaction within the stromal layer, an experimental model of I-CXL is needed. For this purpose our team has proposed, since 2009, the use of an innovative ultrasound system called supersonic shear wave imaging (SSI) for corneal biomechanics imaging. ¹³ A previous study has already demonstrated the feasibility of in vivo corneal stiffness monitoring in a porcine model. ¹⁴ The purpose of the present study was to analyze corneal biomechanical changes induced by the I-CXL protocol using SSI technology with an in vivo rabbit model. 

Methods

Animal Preparation

Six rabbits (New Zealand White albino, males; Charles River Laboratories, L'Arbresle, France) were involved in this study to allow six in vivo I-CXL procedures with six contralateral controls and subsequent ex vivo SSI inflation testing in all 12 eyes. Each rabbit underwent unilateral I-CXL on the right eye; the left eye was used as intra-individual control. All rabbits were treated under general anesthesia performed with ketamine hydrochloride (Virbac, Carros, France) and acepromazine (Calmivet; Vétoquinol, Paris, France). Animals were euthanized after the treatment by intravenous injection of Propofol and both eyes were enucleated for SSI testing. Procedures involving experimental animals were performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and approved by the ethics committee of Bordeaux University of Medicine (Bordeaux, France). 

Supersonic Shear Wave Imaging

Details of the SSI technique can be found in previous publications. ^13,14 Briefly, SSI is an ultrasound-based technique able to map tissue elasticity in real time. This is achieved by quantifying the speed of shear waves V_S (directly linked to elasticity E by the relationship E ≈ 3ρV_S ², where ρ is the density) generated within the tissue. Shear waves are generated by focused ultrasound (focused beams) and detected in real time by using ultrafast ultrasound imaging (up to 20,000 frames/s). The entire sequence is performed by one single linear ultrasound probe (15 MHz central frequency, 128 elements; Vermon, Tours, France) driven by an ultrafast ultrasound scanner (Aixplorer; Supersonic Imagine, Aix-en-Provence, France). ¹³ In the dedicated sequence for the cornea, six “focused beams” are applied successively at equally-spaced locations across the cornea diameter. The whole acquisition sequence lasts a few milliseconds. Each focused beam is used to rebuild a partial elasticity map. Then the six partial reconstructions are recombined to yield a whole cornea map. Then by using a time of flight algorithm on the acquired shear wave movie, the shear wave speed is computed to rebuild an elasticity map. Elasticity maps are then superimposed over the conventional B-mode ultrasound image with a frame rate of 1 map/s, which allows real-time visualization of elasticity changes. 

Iontophoresis Device and I-CXL Protocol

The iontophoretic device consists of a constant current source and two electrodes. The substance to be delivered is applied with a main electrode carrying the same charge as the substance, while a return electrode carrying the opposite charge and placed at another location of the body, completes the circuit. The substance conducts the current through tissues. For rabbits, the return electrode is a small gauge needle placed in the nape of the neck. The main electrode is a circular cup with an internal diameter of 9 mm (Iontofor-CXL; Sooft Italia, Montegiorgio, Italy), surrounded by a 1-mm wide annular suction ring to affix the device on the cornea during the procedure. The electrode itself is a stainless steel grid, placed into the cup at an average distance of approximately 8 mm from the cornea, allowing air bubbles generated by electrolysis of water to escape in order to prevent current supply disruption. The reservoir was filled with riboflavin solution formulated for iontophoresis (Ricrolin; Sooft Italia). Its large volume (0.5 mL) prevents pH and concentration shifts to occur during the application time while using a minimal quantity of buffer that competes with the substance for delivery. The generator applies a constant current of 1 mA for a 5-minute preset period, yielding a delivered dose of 5 mA/min, while controlling the continuity during the procedure and delivering warnings in case of current disruption. 

Once the anterior part of the stroma reaches an adequate concentration of riboflavin (5–10 minutes), the UV-A irradiation can take place and cross-linking of collagen fibers can immediately occur in this area. In that configuration, accelerated UV-A protocol (A-CXL) ¹⁵ is more suitable to experimental models than the conventional UV-A protocol because the shorter exposure time prevents (1) excessive decrease of riboflavin concentration in the anterior part of the cornea during the application time of UV-A, (2) excessive fluctuation of corneal hydration before biomechanical measurements, ¹⁶ and (3) reduces the risk related to anesthesia. 

In our study, two consecutive cycles of 5 minutes were performed for a total of 10 minutes. To avoid any unnecessary strain on the rabbit's eyeball, we used only adhesive strips to maintain subsequent eyelid opening without forceps. Iontophoresis-CXL was performed first while the opposite eye was temporarily closed with strips. The rabbit's anatomy allowed direct and easy access to the corneal suction area, which was successfully obtained on the first attempt in all cases. The integrity of the corneal epithelium was verified before and after I-CXL with a fluorescein solution under blue light testing using our handheld slit-lamp. During iontophoresis, the cornea was immersed under riboflavin fluid, and after iontophoresis, fluorescein-staining test was performed after rinsing the surface with balanced salt solution (BSS) in order to remove the remaining riboflavin. Fluorescein was also removed with rinsing before UV-A irradiation. After the iontophoresis process, yellowing of the corneal tissue was clearly visible in the riboflavin soaked area. The UV-A beam (8-mm diameter, Vega CBM X-linker; Sooft Italia) was immediately focused on this area for 9 minutes with an irradiance of 10 mW/cm² (total dose of 5.4 mJ/cm²; Fig. 1). 

Figure 1

 

(A) Iontophoresis device used with a reservoir filled with riboflavin; (B) illustration of the riboflavin-soaked area (white star) and the riboflavin fluorescence (black star) when the illumination beam is voluntarily displaced.

We did not measure temperature, osmolarity, or pH changes ¹⁷ before or after the iontophoresis process because we were informed by the iontophoresis manufacturer that in the range of our experimental protocol, epithelium integrity, and riboflavin fluid stability were previously established and reported in the safety procedure to obtain Conformité Européenne marking (Roy P, unpublished data, 2011). 

Corneal Elasticity Monitoring Protocol and Inflation Tests

After I-CXL, animals were euthanized and corneas from both eyes were removed with a scleral rim of at least 2 mm. Then, each cornea was immediately mounted on an artificial anterior chamber (AAC; DORC, Zuidland, The Netherlands) (Figs. 2A, 2B). The pressure was continuously monitored with a digital manometer (BiosebAZ 8215; Bioseb, Vitrolles, France) connected to one port of the AAC. A liquid was infused through the second port of the AAC in order to adjust pressure within the chamber. The infusion liquid was a hypertonic solution used for the purpose of deswelling in corneal graft procedures (ThinC; Alchimia, Padova, Italy), the same liquid was used to bathe the anterior surface of the cornea while performing elastography acquisitions. The ultrasound probe was vertically placed above the cornea, perpendicular to the corneal surface. The probe was mounted on micrometer translation and rotation stages allowing very precise three-dimensional (3D) positioning (Fig. 2C). The probe was aligned with the corneal apex using real-time B-mode display. Once alignment was optimized, the shear wave elastography mode was activated, providing real-time visualization of corneal stiffness (kPa) represented by a color scale. Additionally, corneal central thickness (CCT) measurements were performed with ultrasonic pachymeter (CS-100; Tomey, Yokohama, Japan) before and after iontophoresis, after CXL, as well as before and after SSI. 

Figure 2

 

(A) One paired excised corneal buttons. (B) Mounting on the artificial anterior chambers. (C) The ultrasound probe (arrow) is mounted on precision translation and rotation stages.

Inflation tests of the cornea were performed as follows: 

1.  
   The pressure in AAC was raised from 15 to 50 mm Hg by increments of 5 mm Hg.
2.  
   Elasticity maps were recorded at each pressure level after 10 seconds of stabilization. Two inflation cycles (rising pressure) were performed for each cornea but only the second cycle was used for further analysis (the first one was used to ensure that there was no fluid leakage).

Mean elasticity values were obtained through the analysis of the central 3 mm of the cornea. Corneal stiffening versus pressure were plotted for both I-CXL and controls. The slope of each curve was computed by applying linear regression analysis. A comparison of that slope between I-CXL and controls was performed for each rabbit. 

Statistical Analysis

A one-way variance analysis was applied to evaluate the difference between the treated- and the control corneas. A one-way variance analysis gives the P value that several samples are drawn from populations with the same mean. Thus, if a low P value is obtained, the null hypothesis is rejected (i.e., the samples are significantly different). Differences between treated groups and control groups were considered significant at P less than 0.05 (Statplus, Microsoft Excel 2008; Microsoft, Redmond, WA). 

Results

Iontophoresis Procedure

Corneal yellowing of the central 9-mm diameter area was clearly visible in the eyes treated using iontophoresis. When exposed to UV-A light, a green fluorescence appeared where the riboflavin had successfully impregnated the cornea. The green fluorescence intensity in this area was not homogenous and exhibited a ring of hyperfluorescence at the edges of the iontophoresis area (outer diameter of 9 mm) in all cases. Furthermore, some “fluorescence hot spots” were also observed in the central area without a specific pattern (diameter of approximately 1–3 mm; Fig. 3). Fluorescein staining with blue light testing was systematically negative before and after I-CXL. 

Figure 3

 

Illustration showing the heterogeneous pattern of riboflavin fluorescence with paracentral hot spots (arrows in [A, B]) and peripheral reinforcement ring (stars in [C, D]) during UV-A period of several I-CXL procedures. (A, B) pictures were taken with top incidence with the Vega CBM X-linker camera, (C, D) are white illumination pictures taken with a lateral incidence.

Corneal Central Pachymetry Monitoring

The CCT was neither significantly affected by iontophoresis nor by CXL. The mean CCT was 388 μm (SD 22 μm) before iontophoresis, 395 μm (SD 21 μm) after iontophoresis, 376 μm (SD 18 μm) after UV-A, 400 μm (SD 18 μm) before SSI, and 392 μm (SD 16 μm) after SSI. Mean CCT for controls were 392 μm (SD 12 μm) before euthanasia, 393 μm (SD 6 μm) before SSI, and 400 μm (SD 11 μm) after SSI. P values comparing in vivo (pre-iontophoresis) and ex vivo (post SSI) for right and left eyes were 0.58 and 0.31, respectively (Table 1). 

Table 1

 

Corneal Central Thickness Measured With an Ultrasound Pachymeter at Different Steps of the Experimental Procedure for Right (R) and Left Eyes (L)

Rabbits' CCT, μm	CCT Pre IONTO	CCT Post IONTO	CCT Post UV-A	CCT Prior SSI (Ex Vivo)	CCT After SSI (Ex Vivo)	CCT Difference R-L Prior SSI (Ex Vivo)	CCT Sifference R-L After SSI (Ex Vivo)
P values (R vs. L)	0.52			0.43	0.91
1	R 375	R 390	R 385	R 410	R 400	25	5 (−)
	L 395			L 385	L 405
2	R 390	R 420	R 365	R 395	R 380	5	25 (−)
	L 380			L 390	L 405
3	R 355	R 375	R 380	R 405	R 410	10	30 (−)
	L 380			L 395	L 380
4	R 390	R 410	R 400	R 425	R 385	35	15 (−)
	L 395			L 390	L 400
5	R 400	R 370	R 350	R 400	R 370	5 (−)	15 (−)
	L 415			L 405	L 395
6	R 420	R 410	R 380	R 370	R 410	25 (−)	5 (−)
	L 390			L 395	L 415
Mean	R 388	R 395	R 376	R 400	R 392	17	15
(SD)	−22	−21	−18	−18	−16	−12	−10
Mean	L 392			L 393	L 400
(SD)	−12			−6	−11

 Corneal central thickness values were rounded to 5-increment units during measurements. P values comparing mean values of right and left eyes are reported. IONTO, iontophoresis.

Corneal Inflation Testing

Figure 4 shows an example of the corneal elasticity imaging during an inflation test. Corneal stiffness increased with increasing pressure in both experimental and control eyes, as expected. A threshold was applied to the elasticity display (color scale) to account for the quality of the elasticity estimation. The area reaching this threshold shrinks as the pressure increases. Indeed, the higher the pressure, the stiffer and less deformable is the cornea, leading to lower amplitudes of the shear wave. This affects the elasticity estimation especially in the peripheral part of the cornea that undergoes a higher stress than the central part. We used a threshold on the elasticity estimation quality to ensure the accuracy and robustness of our measurements. 

Figure 4

 

Illustration of one ex vivo inflation testing showing corneal elasticity imaging (color scale, kPa) at 35 mm Hg of pressure. The red dots superimposed to the elasticity maps indicate the lateral positions of the “pushing beams.” There is a 2-mm spacing between two consecutive red dots.

For all the rabbits examined in our study, increasing pressure revealed differences between corneas that underwent I-CXL and their respective controls. In the relevant range of stress that allows subsequent SSI imaging (inflation test from 15–50 mm Hg), the elasticity variation appeared to be a linear function of pressure: E = αp+β, where E is Young's modulus (kPa), p the inflation pressure (mm Hg), α and β coefficients of linear regression, the slope and intercept, respectively (Table 2). Figure 5 shows the slopes of both treated- and control eyes for each rabbit elasticity-pressure curve (n = 6). The mean slope was 27.1 and 16.9 kPa/mm Hg, respectively, in the I-CXL group and control group, reaching a statistically significant increase factor of a 1.6 (P = 0.029) with one tailed t-test. 

Figure 5

 

Inflation testing with correlations of elasticity and inflation on the right I-CXL cornea compared with the left untreated cornea for the rabbits 1 through 6. The fitting curves and their linear equations serve to quantify differences between groups. See corresponding P values for each level of pressure in Table 3.

Table 2

 

Alpha and β Coefficients of the Linear Regression Curves (E = αp + β) and Corresponding Values From Right Eye Minus Left Eye

Rabbits	α Right	α Left	α variation R–L	β Right	β Left	β variation R–L
1	20.4	11	9.4	0.6	14	−13.4
2	21.4	10.6	10.8	58.8	14.9	43.9
3	40	28.7	12.7	34.5	14	20.5
4	16.9	10.1	6.8	64.2	64.3	−0.1
5	30.1	21.5	9.6	8.2	18.4	−9.8
6	33.7	19.4	14.3	26.7	17.7	9
Mean (SD)	27.1 (8.9)	16.9 (7.5)	10.2 (2.6)	32.2 (25.8)	23.9 (19.8)	8.4 (21.3)

Table 3

 

Young's Modulus Mean Values Versus Different Load of Pressure and Corresponding P Values (n = 6) in Right (R: I-CXL Group) and Left Eyes (L: Control Group)

Load mm Hg	Mean E (SD) kPa R	Mean E (SD) kPa L	P ANOVA Test
15	36 (28)	34 (25)	0.9326
20	62 (34)	47 (18)	0.3629
25	93 (34)	66 (19)	0.1324
30	124 (30)	95 (33)	0.1200
35	151 (39)	108 (35)	0.0790
40	178 (41)	121 (41)	0.0395
45	200 (38)	139 (41)	0.0262
50	222 (31)	145 (37)	0.0034
Mean	133 (34)	95 (32)	–

 Bold values are statistically significant (P < 0.05).

Mean elasticity values versus inflation testing were summarized in Table 3 for each level of pressure. Mean differences between groups were statistically significant for pressures above 40 mm Hg (P < 0.05). 

Discussion

Corneal stiffening due to the CXL is a subtle and challenging interaction to detect in clinical practice and, to date, there is no reliable tool to assess variations of the corneal stiffness in patients after CXL. ¹⁸ One of the main reasons is probably the difficulty to distinguish stiffness changes related to CXL with stiffness changes related to in vivo IOP variations. Laplace's Law for a pressurized vessel states that wall tension is a function of the internal pressure, such that the greater the internal pressure, the greater the wall tension. This means that as IOP increases, the cornea becomes stiffer due to its known nonlinear properties. Another important concern is to get reliable physical parameters corresponding to the tissue stiffness instead of indirect measurements that are viscoelastic in nature (e.g., corneal hysteresis). Thus, in order to demonstrate and to quantify the stiffening induced by any CXL protocol, it remains necessary to get (1) an experimental model with controlled parameters (i.e., hydration state, corneal shape and pachymetry, and load of pressure and motion), (2) an appropriate elastometry technology that is as minimally invasive as possible (i.e., remote observation, minimal stress, geometrical and histological homeostasis, and respect of the boundary conditions). We will discuss successively the choice of our experimental setup, the main results we obtained with I-CXL, the study limitations and further work to be done. 

Experimental Setup and Model Justification

Rabbits eyes have many similarities with human eyes, but also important differences that are biomechanically relevant: very soft eyeball, smaller corneal thickness (∼400 μm), larger corneal diameter (13 ± 0.30 mm), and greater corneal curvature (radius 7.26 ± 0.26 mm). ^19,20 In addition, the rabbit cornea does not have any Bowman's membrane, which is a possible additional barrier to the permeability of riboflavin during CXL protocols. 

We first attempted to perform in vivo elasticity monitoring of the I-CXL procedure. The position of the rabbit body, lying on its side, was found to be the best position to limit breathing movements during in vivo CXL and SSI examination. Nevertheless, in the rabbit anatomy, there is no bone frame to protect the eyeball from external trauma. In our experience, when one of the eyes was aligned vertically toward the elastography probe, the opposite eye was directly compressed on the examination table. From our preliminary tests performed in this position, we observed a significant softening of cornea within a few minutes, probably because the IOP was dramatically decreased. Thus, only the first eye examined could be considered reliable for in vivo pre- and post-CXL comparison. Another confounding factor was the anesthesia depth that could induce unpredictable variations of the IOP (blood pressure, eyelid tonicity) and consequently undesirable changes of the corneal stiffness. Finally, after many trials to overcome the in vivo IOP variations, our conclusion was that the CXL had to be performed in vivo, but the elasticity measurements with SSI required very accurate IOP monitoring, which could only be performed in the ex vivo situation. Therefore, we decided to mount the rabbit's corneas on an AAC immediately after euthanasia, and to use a manometer for adequate inflation pressure control. In order to isolate the effect of the photopolymerization process itself from any other biological secondary process, animals were killed immediately at the end of the I-CXL procedure. Biological secondary effects are expected to vary from human to rabbit species. 

Finally, since our biomechanical measurements were performed in the ex vivo condition, the in vivo translation of our conclusions must be done with caution. Possible changes of corneal hydration state could modify the biomechanical behavior. Furthermore, enzymatic degradation processes could occur during the hours after the rabbit's euthanasia with a possible postmortem corneal softening. We attempted to minimize the effects of these postmortem changes on our conclusions by using the experimental design of treated and control eyes from the same animal. 

Biomechanical Changes After I-CXL: Pressure-Inflation Testing

As previously mentioned, the corneal stiffness is a function of the biomechanical load, which is represented by the pressure in the AAC. This statement is consistent with knowledge of the ultrastructural properties of corneal tissue. The physical load of pressure will stress the collagen fibers and change their 3D structure, depending of the interconnections of the collagen fibers and the surrounding fundamental substance. It was already demonstrated by Tan et al. ²¹ that the “wavy structure” of the collagen could be increased after CXL. In the current study, corneas that underwent I-CXL exhibited different biomechanical inflation profiles. The stiffness differences were not clearly visible below the 25 mm Hg load of pressure inflation, but became obvious between 25 and 50 mm Hg and statistically significant above 40 mm Hg. This behavior might suggest a possible change of the biomechanical properties of corneas induced by I-CXL, in a nonlinear fashion. 

The relationship between elasticity and pressure was fit with a linear equation. The slope of this linear function was consistently higher in the I-CXL cornea compared with the contralateral untreated cornea in all cases, highlighting stiffening in the corneal biomechanical properties. Even though the comparison of mean Young's modulus values at individual pressure levels below 40 mm Hg did not produce a significant difference between treated- and control corneas, the increased slope of the change in Young's modulus over a wide range of pressures in treated versus control corneas leads to the conclusion of stiffening. In other words, I-CXL induced greater change in stiffening than in control groups with increasing pressure load. To our knowledge, no anterior data are available in the literature regarding transepithelial inflation testing, even though results of quasistatic optical coherence tomography (OCT) elastography in TE-CXL (BAK) model have shown some similarities with ours. ²²  

Finally, we do not have enough data to conclude whether iontophoresis-induced riboflavin pooling might occur in the epithelium. This situation could shield the UV-A light and limit stromal CXL. Nevertheless, electroporation (transcellular pathway) of riboflavin during iontophoresis (intercellular pathway) is probably not a reasonable issue according to a previous study reported in the literature. ²³ Furthermore, it has already been reported for C-CXL with transepithelial riboflavin soaking (with riboflavin penetration enhancers) that the epithelium could play a significant role in blocking stromal absorption of UV-A. ²⁴ Nevertheless, I-CXL is a new procedure, and eventually the UV-A dose could be adjusted to account for epithelial absorption after adequate investigation of the effect. 

Study Limitations

In terms of imaging, when the pressure is increased, there is a spatial reduction in the area of elasticity mapping of the cornea. This can be explained by at least two different physical phenomena. (1) As previously mentioned, elasticity values are only displayed if the quality of the estimate reaches a certain threshold: only a correlation coefficient greater than 0.9 are considered when applying the time-of-flight correlation algorithm. This quality is affected by the amplitude of shear wave compared with the noise of shear wave detection (signal-to-noise ratio). As pressure increases, the cornea becomes stiffer and less deformable. Thus, the shear wave exhibits lower amplitude. Therefore, we observe a decrease in the available area of analysis, and (2) as demonstrated by Nguyen et al., ²⁵ the cornea can be considered as an immersed plate. In such geometrical conditions shear waves are guided into the cornea and their attenuation is mostly due to thickness and curvature. Thus, as pressure increases, the curvature increases and the ultrasound beams become more and more parallel to the surface of cornea, leading to a less effective acoustic pressure generating shear waves. Moreover, the current time-of-flight algorithm (used to calculate the shear wave speed) uses directional filters that track the shear wave perpendicularly to the ultrasound axis. As the shear wave is guided by the cornea geometry, when the curvature increases, shear waves propagate more and more parallel to the ultrasound axis. One should notice that these artifacts could be solved in a future refined version of the algorithm for shear wave estimation. 

The corneal thickness can also play a role in the conversion of shear wave speed into elasticity. The cornea acts as a wave-guide due to its geometry: both the posterior and the anterior boundary of cornea constrain the shear wave propagation to a particular mode (“guided mode”) corresponding to the case of Lamb-like waves. According to the Lamb waves equations, the thinner the cornea is, the more constrained and slowed the shear wave will be. ²⁵ Thus, a thinner cornea would appear softer. Unfortunately, the ultrasound B-mode axial resolution was only 150 μm with the 15-MHz probe we used, limiting our analysis of small variations of the corneal thickness. However, we carefully monitored the CCT throughout the whole procedure using an ultrasound pachymeter (30-μm axial resolution). No significant difference was found between the left and right eyes of the rabbits, thus the comparison of their relative Young's moduli is still valid. Monitoring the corneal thickness was also a key point to ensure that elastic property changes were not induced by excessive hydration variations. ²⁶ According to Table 1, the comparison of the CCT of right versus left corneas, and I-CXL versus controls, did not show any significant change throughout the different steps of our experimental procedure. The deswelling liquid that we used for bathing the cornea on both sides probably helped in this situation. 

Very few papers are available about I-CXL protocols, but the feasibility in humans has already been demonstrated. ²⁷ In our study, the heterogeneous fluorescence of the riboflavin after iontophoresis raised some questions. In theory, if both electric field and tissue permeability were homogenous, riboflavin concentration in the stroma should be constant. In other words, unless we were not measuring riboflavin concentration in the corneal tissue, the riboflavin fluorescence should be homogeneous across the iontophoresis area. That was not the case for our rabbits. In all the cases, we observed small fluorescence “hot spots” (1- to 3-mm wide) randomly visible during UV-A irradiation, inside the iontophoresis area and associated with a ring of higher fluorescence located at the edges of this iontophoresis area (outer diameter of 9 mm). We can assume that stretching forces applied on the epithelium layer during the application of the suction ring could explain this peripheral riboflavin hyperfluorescence. Nevertheless, hot spots are more difficult to explain and could be correlated with some local changes of the corneal epithelium permeability to riboflavin. This phenomenon was not observed in preliminary trials on human corneas, ²⁷ but the rabbit model is distinct from the human because, histologically, there is no visible basal membrane (e.g., Bowman's layer) to ensure a homogenous anchoring of the epithelium into the stroma. Furthermore, experimental rabbits are usually very young (2–3 kg and a few months old), thus their epithelium might be very cohesive with the underneath stromal layer compared with adult human corneas. We did not find any contributive publication on this topic. 

We looked at previous experimental studies using corneal iontophoresis with testing on rabbits or rats, including histology that did not report epithelial injuries. ^28–30 Furthermore, we do not believe that observed hot spots are correlated with artifact or epithelial injury. Rabbits were managed with considerable caution, avoiding the use of a speculum, corneal exposure, and desiccation. The iontophoresis device when carefully applied, does not injure the epithelium except possibly at the level of the suction ring near the corneal periphery. Corneal pachymetry measurements with the ultrasound probe were always performed in the center, with as little contact as possible. Direct observation with handheld slit-lamp did not exhibit any visible corneal fluorescein staining. 

Nevertheless, it is interesting to note that, since fluorescein and riboflavin can both absorb light in the blue region of the spectrum (435–465 nm) ^31,32 and will emit green light (to 500–550 nm), ^33,34 observation of the epithelial integrity might be difficult after iontophoresis and could represent, retrospectively, a limitation of our study. Thus, minor injuries of the epithelium layer could be present despite the negative corneal staining. Epithelial mapping with OCT or high-resolution echography could help to understand the phenomenon of riboflavin hot spots observed in rabbits. Thus far, information concerning iontophoresis safety, previously mentioned, were only obtained from corneal tangential cuts with conventional histology and not en face. ^28–30  

Next Steps and Questioning for the CXL Quest for Evaluation

Our results provide experimental evidence of the transepithelial (via iontophoresis) stiffening by the combined action of UV-A light with riboflavin in an in vivo rabbit model. However, it remains unknown how long this stiffening effect can be maintained. In addition, we do not know if the dominant mechanism of the clinical efficacy of CXL is an architectural reinforcement of the cornea by the way of instantaneous collagen fibers polymerization (corneal stiffening) or rather a long-term biological healing process occurring within months after CXL (epithelial and superficial collagen fiber anchoring renewal, new arrangements of the keratocyte network after keratocyte loss, renew of superficial nervous plexus in “epithelium off” CXL, etc.). ^35,36 In that way, TE-CXL evaluation in subphysiologic conditions is useful since epithelial healing cannot be considered as a confounding factor for clinical demonstration of photopolymerization in CXL efficacy. It is interesting to mention that this instantaneous photopolymerization effect was, from the beginning of CXL experimental trials, the main rationale to explain CXL clinical efficacy. ³⁷ To our knowledge, only two papers reported extensiometry measurements after C-CXL in a relative long-term, follow-up period. The first one was performed by Spörl et al. ³⁸ in 16 rabbits' eyes with timeline euthanizations at 1 month (n = 11) and 3 months (n = 8). The second one was published by Wollensak et al. ³⁹ in nine rabbits at 1, 2, and 3 months (n = 3 for each period). Recently, Ford et al. ⁴⁰ described a corneal elastography technique based on speckle tracking with OCT that could be possible for future in vivo CXL longitudinal studies. Many studies about microstructural changes after CXL have been published, but the correlation between structural changes and biomechanical properties were neither clearly established nor conducted in the in vivo situation. ^21,41,42 Unfortunately, to date, there is no way to clinically follow collagen architectural changes with time after CXL in living patients. With promising approaches to image the effective in vivo microstructural organization of the corneal collagen network (e.g., second-harmonic generation imaging), the field of in vivo corneal elastography research remains a very important path for understanding the science of CXL. 

For these reasons, we believe that SSI technology might be one of the most relevant options for experimental studies in the scope of corneal biomechanics imaging with a realistic hope for future clinical use. Thus, one interesting practical option for in vivo human applications could be to draw the individual corneal elasticity behavior under different load of pressure with the featuring of different parameters such as the slope and the nonlinearity of the rising elasticity values. Furthermore, the corneal elastic anisotropy induced by CXL could be an interesting biomechanical parameter to evaluate. Finally, to overcome present technologic limitations, upcoming developments will aim to upgrade the probe frequency and to couple our system with OCT. 

Conclusions

Collagen cross-linking is a complex and multifactorial process where biomechanical changes are expected to be a highly valuable indicator related to the associated clinical efficacy in retarding keratoconus progression. Our study highlights biomechanical changes induced by I-CXL under precisely controlled conditions (pressure, hydration, and thickness). As a main result, corneas treated with I-CXL exhibited a sharper stiffening under pressure rise than control corneas. More extensive studies are needed in the future to further evaluate this relationship. We think that SSI monitoring could be considered as a valuable option for the independent assessment and titration of the photopolymerization process occurring with in vivo TE-CXL. 

Acknowledgments

The authors thank Institut national de la santé et de la recherche médicale (INSERM), 1034 unit (Thierry Couffinhal, MD), and the Département de Techniques et de Recherches Chirurgicales Appliquées unit (Jean Claude Le Huec, MD, PhD) for their help with animal housing and technical support. Also, Pierre Roy, PhD, for his support with iontophoresis technology understanding. 

Supported by grants from the French National Agency for the Research (ANR Microelasto) for the funding of this project. 

Disclosure: D. Touboul, None; J.-L. Gennisson, None; T.-M. Nguyen, None; A. Robinet, None; C.J. Roberts, None; M. Tanter, None; N. Grenier, None