Monitoring of Cornea Elastic Properties Changes during UV-A/Riboflavin-Induced Corneal Collagen Cross-Linking using Supersonic Shear Wave Imaging: A Pilot Study

Abstract

Purpose.: Keratoconus disease or post-LASIK corneal ectasia are increasingly treated using UV-A/riboflavin–induced corneal collagen cross-linking (CXL). However, this treatment suffers from a lack of techniques to provide an assessment in real-time of the CXL effects. Here, we investigated the potential interest of corneal elasticity as a biomarker of the efficacy of this treatment.

Methods.: For this purpose, supersonic shear wave imaging (SSI) was performed both ex vivo and in vivo on porcine eyes before and after CXL. Based on ultrasonic scanners providing ultrafast frame rates (∼30 kHz), the SSI technique generates and tracks the propagation of shear waves in tissues. It provides two- and three-dimensional (2-D and 3-D) quantitative maps of the corneal elasticity.

Results.: After CXL, quantitative maps of corneal stiffness clearly depicted the cross-linked area with a typical 200-μm lateral resolution. The CXL resulted in a 56% ± 15% increase of the shear wave speed for corneas treated in vivo (n = 4).

Conclusions.: The in vivo CXL experiments performed on pigs demonstrated that the quantitative estimation of local stiffness and the 2-D elastic maps of the corneal surface provide an efficient way to monitor the local efficacy of corneal cross-linking.

Introduction

UV-A/riboflavin–induced corneal collagen cross-linking (CXL) has been proposed over the last decade as a treatment for keratoconus. ^1,2 This minimally invasive technique is expected to stop the progression of keratoconus and hence to avoid or postpone corneal keratoplasty. It consists in photo-reticulating the collagen fibers of the cornea in order to stiffen the corneal structure. The cornea is soaked with a photosensitizer (riboflavin) and then exposed to UV-A illumination. Ex vivo tensile tests on cornea strips were performed to estimate the elasticity variations induced by CXL. ³ Clinical trials have been conducted in vivo. ^4,5 The in vivo efficiency of the CXL is usually estimated by measuring different parameters such as the patient's visual acuity, corneal topography, and corneal keratometry over a few years after the treatment every 3 or 6 months. 

This therapeutic approach suffers from a lack of techniques to provide an assessment of the biomechanical effects of the treatment. Whereas current indicators are focused on morphologic measurements, the use of mechanical properties as an in vivo indicator of treatment efficiency still needs to be investigated. The ocular response analyzer (ORA; Reichert, Depew, NY) was proposed to correlate the corneal elasticity with the corneal hysteretic response to applanation forces. ⁶ But this technique failed to demonstrate significant changes after CXL. ⁷ In this study, we propose supersonic shear wave imaging elastography (SSI) as a real-time and noninvasive ultrasound-based method to assess the corneal elasticity variations induced by CXL. Elastography is an emergent technique for medical diagnosis. Various techniques have been developed to recover the viscoelastic properties of a biological tissue by measuring its response to a mechanical stimulus. In the case of shear wave imaging, ⁸ the stimulus is the ultrasonic radiation force, which induces a transient shear wave propagating in the tissue. The shear wave propagation speed is linked to the shear modulus of the tissue. The particularity of the SSI method lies in the use of an ultrafast ultrasound scanner to image the shear wave propagation at a very high frame rate (up to 30,000 frames per second) in order to measure its speed. The ability of SSI elastography technique to assess the biomechanical properties of the cornea has been demonstrated in previous studies. ⁹ Here, CXL experiments combined with SSI elastography monitoring were conducted ex vivo and in vivo on porcine eyes. 

Materials and Methods

Corneal Collagen Cross-Linking

UV-A/riboflavin–induced CXL has been introduced in the last decade as an in vivo cornea stiffening method. Cross-links between the collagen fibrils can be induced by photo-polymerization, leading to a stiffer fiber network. 

The CXL protocol is based on three steps. First, the epithelium is removed from the central part of the cornea so that the cornea becomes permeable. Then, riboflavin solution (riboflavin 0.1% dextran T500 20%; Horus Pharma, Saint-Laurent du Var, France) is dropped on the cornea (1 drop/min for 30 minutes) in order to enable the corneal stroma to absorb the UV-light emission without damaging the deeper structures (endothelium, lens, retina). Finally, the cornea is exposed to UV-light–emitting diodes (wavelength 370 nm, intensity 3 mW/cm²; IROC Medical, Zurich, Switzerland) for 30 minutes while riboflavin drop is continued (1 drop every 5 minutes). In our experiments, the diameter of the irradiated area was set to 8 mm. 

Supersonic Shear Wave Imaging

Young's modulus (E) of a homogeneous tissue can be retrieved from the speed (c_T ) of a shear wave propagating in this tissue: 

 

where ρ is the tissue density. 

The SSI is an ultrasound-based elastography technique based on Equation 1. It is a combination of the generation of a plane shear wave into the tissue using the ultrasonic radiation force and of ultrafast ultrasonic imaging to the propagation of the shear wave in the tissue. 

An ultrafast scanner (Aixplorer; SuperSonic Imagine, Aix-en-Provence, France) is used to drive a standard ultrasonic probe (linear array, 128 elements, 15 MHz) by per-channel programming in both transmit and receive. First, the ultrasonic beam is focused for a few tens of microseconds in the tissue to induce a transient axial displacement (Fig. 1a). The tissue relaxation generates a shear wave, which is propagating perpendicularly to the ultrasonic pushing beam axis. Then the probe is switched to the ultrafast imaging mode: the tissue is imaged at a very high frame rate (up to 30,000 frames per second) during the shear wave propagation (Fig. 1b). These two steps are repeated at different locations in order to map the entire cornea. As illustrated on Figure 1c, four pushing beams are applied at different abscissae along the cornea curvature. Each pushing beam is followed by an ultrafast imaging sequence. 

Figure 1. 

 

Supersonic shear wave imaging principle. (a) First, a shear wave is induced in the cornea using the ultrasonic radiation force. (b) Second, the probe is switched to an ultrafast imaging mode (up to 30,000 frames per second) in order to follow the shear wave propagation. (c) In order to map the entire cornea, four pushing beams are applied successively at different locations. The pushing beams are interleaved with ultrafast imaging sequences.

Figure 2 shows the axial displacements that are induced by the ultrasonic radiation force in the cornea. The magnitude of these displacements is typically a few micrometers. The shear wave speed is deduced by cross-correlating the images of the displacement field. 

Figure 2. 

 

Snapshots of the shear wave propagation in a porcine cornea. The ultrasonic radiation force is applied at t = 0 ms at the location indicated by the white arrow. The resulting axial displacements are superimposed to the echographic image. The color scale corresponds to the amplitude of these displacements (red, toward the probe; blue, downward).

In the case of layered organs, the shear wave is guided along the layer by the successive reflections on the layer boundaries. This guided wave phenomenon results in dispersive effects that must be taken into account to recover the shear modulus. As shown in previous studies, ^9,10 the dispersion curve can be approximated by 

 

where v_φ is the phase velocity of the shear wave, ω is the angular frequency of the shear wave, h is the layer thickness, and c_T is the intrinsic shear wave speed of the organ. 

The phase velocity is defined as the speed of a single-frequency component of the shear wave. In our experimental configuration, the 15-MHz ultrasonic pushing beam induces a broadband shear wave (typically 200–1500 Hz). In this article, we express the results in terms of group velocity, which corresponds to the speed of the whole shear wave packet. 

Statistical Analysis

A statistical analysis was performed on the results of the ex vivo CXL experiments. A one-way ANOVA was applied to differentiate the successful treatments from the unsuccessful ones. ANOVA gives the probability (P value) of the null hypothesis 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). 

Ex Vivo Corneal Cross-Linking Experiments.

We first tested our elastography monitoring protocol on ex vivo eyes. Freshly excised porcine eyeballs were obtained from a slaughterhouse (Etablissements Guy Harang, Paris, France) a few hours postmortem. The eyeballs were kept refrigerated in physiological saline solution. The experiments were performed within 48 hours after enucleation. 

In order to be close to physiological conditions, we performed a monitoring of the IOP. A needle was inserted in the anterior chamber and linked to a digital manometer (AZ 8215; Bioseb, Vitrolles, France). Physiological saline solution was injected in the anterior chamber to adjust and maintain the IOP to 20 ± 0.4 mm Hg. 

For n = 9 eyes, elastography acquisitions were performed before the treatment and after de-epithelization and riboflavin application (control group). For n = 17 eyes, elastography acquisitions were performed before and after a complete CXL procedure (de-epithelization, riboflavin application, and UV exposure). 

During the elastography sequences, the eyeballs were immersed in a water tank, and the probe was placed above the cornea as shown in Figure 3. During the riboflavin instillation and the UV-A–exposure steps, water was pumped out of the tank and the ultrasonic probe was translated upward without changing the eyeball position. 

Figure 3. 

 

Experimental setup for ex vivo SSI acquisitions. The eyeball is immersed in water. A needle is inserted in the anterior chamber to regulate the IOP. The ultrasonic probe (15 MHz, 128 elements) is placed a few millimeters above the cornea.

In Vivo Corneal Cross-Linking Experiments.

In vivo experiments were conducted on porcine eyes (n = 4) at the Institut Mutualiste Montsouris Research unit (IMM Recherche, Paris, France) after validation by the local ethics committee. The pigs were under general anesthetic and mechanical ventilation. All the animals were treated in accordance with the ARVO statement for the use of animals for ophthalmic and vision research. 

The animal eyelids were maintained open using a palpebral speculum. A rubber ring was used to form a small hermetic water tank around the animal ocular orbit. The ring was strapped on the head of the pig. The ultrasonic probe was mounted on an articulated arm. Elastography acquisitions were performed before and after the CXL treatment. During the elastography acquisitions, the tank was filled with physiological saline solution, and the probe was placed above the cornea. The in vivo experimental setup is shown in Figure 4. 

Figure 4. 

 

Experimental setup for in vivo acquisitions on anesthetized pigs. (a) A rubber ring is placed around the eye to enable the cornea immersion during elastography acquisition. (b) Riboflavin 0.1% is dropped on the cornea for 20 minutes. (c) The cornea is exposed to UV-A light for 30 minutes.

The central corneal thickness was measured before and after CXL using a pachymeter (SP 100; Tomey, Nagoya, Japan). We minimized as much as possible the undesirable motions of the cornea: the head was strapped to the operating table; and we triggered the elastography acquisitions with both the respiratory cycle and the cardiac cycle, in order to avoid global displacements caused by the respiration or the pulsatility. 

After the in vivo experiments, the animal was euthanized, and the eyeballs were collected. Ex vivo SSI acquisitions were then performed on the enucleated eyeballs. We performed three-dimensional (3-D) scans of the cornea by rotating the probe above the cornea using a dedicated mounting. Elastic maps of the corneal surface were reconstructed from acquisitions in different imaging planes (from 0° to 180° by steps of 5°; i.e., 37 imaging planes). 

Results

Ex Vivo CXL

Figure 5 shows an example of the images that can be obtained on an intact enucleated eye using a 15-MHz linear array. The intact ex vivo cornea exhibits a homogeneous shear wave speed (2.0 ± 0.2 m/s). 

Figure 5. 

 

(a) B-mode image of an ex vivo porcine cornea acquired with a 15-MHz linear array. (b) Elastic map of an ex vivo porcine cornea, obtained using SSI, superimposed on the B-mode image. The color scale corresponds to the shear wave group velocity (m/s).

CXL was applied ex vivo on a total of 17 eyes. For these treated eyes, the elasticity was measured before and after the CXL. A control group was composed of 9 eyes for which the elasticity was measured before and after both epithelium removal and riboflavin application. For the control group, no significant change of the corneal elasticity was observed (variation of −2% ± 15% of the shear wave speed). For five of the treated eyes, significant corneal stiffening after CXL was observed (increase of 60% ± 20% of the shear wave speed). On the other corneas, the elasticity did not increase after treatment and even decreased in some cases (shear wave speed variation of −20% ± 20%). 

Figure 6 shows the shear wave speed variations obtained in the responsive group (n = 5 over 17 treated eyes) compared with the control group (n = 9). Both groups are significantly different (P < 4.10⁻⁵). This P value was obtained using one-way ANOVA to validate the discrimination between both groups. 

Figure 6. 

 

Elasticity increase induced by CXL corneas. The control group was de-epithelized and soaked with riboflavin but not exposed to UV irradiation (n = 9). The (+) group corresponds to corneas that were responsive to the CXL treatment (n = 5). Both groups are significantly different (P < 4.10⁻⁵).

In Vivo CXL

To avoid ex vivo biases, we performed CXL in in vivo conditions on anesthetized pigs. A total of n = 4 porcine eyes were treated in vivo. 

Figure 7 shows the elastic maps obtained on a pair of eyes belonging to the same animal after enucleation. The values of the shear wave speed are superimposed on the echographic images. CXL was applied in vivo on the left eye, while the right eye remained untreated and was used as a control. The control cornea (Fig. 7b) exhibits a homogeneous elasticity (2.0 ± 0.2 m/s). On the treated cornea (Fig. 7a), two areas can be distinguished. The central part (delineated by the white dashed line), corresponding to the UV-A–beam localization, is stiffer (3.4 ± 0.5 m/s) than the peripheral part (2.1 ± 0.1 m/s). Furthermore, the elasticity in the peripheral part (not exposed to UV-A–illumination) is close to the elasticity of the control cornea. 

Figure 7. 

 

Ex vivo elastic maps acquired on porcine corneas superimposed on the echographic image. The color scale corresponds to the shear wave group velocity (m/s). A CXL was performed in vivo on the central part (delineated by the white dashed line) of the left cornea (a), while the right eye (b) of the same animal remained untreated. The color scale corresponds to the shear wave group velocity (m/s).

The results obtained on another porcine cornea are presented in Figure 8. As illustrated in Figure 8a, the epithelium was removed only on the superior half of the cornea. During the UV-A–exposure, a higher light intensity was observed on the de-epithelized area than on the epithelized area (Fig. 8b). The riboflavin appears with a fluorescent-green color when illuminated by UV-A light; the more riboflavin absorbed, the more intense the green color. Thus, we can conclude that the riboflavin is mainly absorbed in the de-epithelized area. 

Figure 8. 

 

(a) In vivo porcine cornea after unilateral epithelium removal (superior half of the central 8 mm of the cornea). (b) In vivo porcine cornea during the UV-A exposure (8-mm diameter beam). (c) Elastic map of one cross-section of the cornea acquired in vivo after CXL. (d) Elastic map acquired ex vivo after enucleation on the same cornea. Anatomic marks are reported on each figure (S, superior; I, inferior; T, temporal; and N, nasal). The imaging plane shown in (c) and (d) is aligned along the inferior–superior meridian. The color scale in (c) and (d) corresponds to the shear wave group velocity (m/s).

Figure 8c shows an elastic map of the cornea acquired in vivo after the CXL. The imaging plane is aligned along the inferior–superior meridian of the cornea. In this cross-section, the right part (S indicates superior) is stiffer than the left part (I indicates inferior): the shear wave speed median values are 6.7 ± 0.3 m/s and 5.1 ± 0.2 m/s, respectively. The SD corresponds to the spatial heterogeneity. 

Figure 8d shows an elastic map of the same cornea acquired ex vivo on the enucleated eyeball a few hours post mortem. The shear wave speed is globally lower than in the in vivo cornea because the IOP is much lower in ex vivo conditions. Again, the imaging plane is aligned along the inferior–superior meridian of the cornea. The shear wave speed median values are 2.1 ± 0.1 m/s in the right part (superior side) and 1.7 ± 0.1 m/s in the left part (inferior side). 

We obtained an ex vivo elastic map of the surface of the cornea from the acquisition of different imaging planes (Fig. 9). For each point, the shear wave speed was averaged over the whole corneal thickness. On this map, the superior half of the cornea appears significantly stiffer (2.17 ± 0.05 m/s) than the inferior half (1.60 ± 0.02 m/s). The geometrical position of the stiffer area matches the de-epithelized area. 

Figure 9. 

 

Ex vivo elastic map of the surface of the cornea after a CXL had been performed in vivo on the superior half on the cornea. The color scale corresponds to the shear group velocity (m/s). For each point, the shear wave speed has been averaged on the whole cornea thickness.

The results obtained on n = 4 corneas unilaterally treated are summarized in the Table. For these four cases, the shear wave speed difference between the treated area and the untreated area was 56% ± 15% (the SD corresponds to the variability between the different experiments). 

Table. 

 

Results Obtained for n = 4 Corneas Treated In Vivo by CXL*

No.	In Vivo CCT (μm), SD = ±50 μm		Untreated Area		Treated Area		vg Variation (%)	E Variation (%)
			vg (m/s), Median ± SD	E (kPa), Median ± SD	vg (m/s), Median ± SD	E (kPa), Median ± SD
	Initial	Final
1	618	614	2.1 ± 0.1	71 ± 14	3.4 ± 0.5	487 ± 287	62	586
2	554	500	1.7 ± 0.2	30 ± 14	2.9 ± 0.3	258 ± 107	70	760
3	782	724	2.00 ± 0.02	58 ± 2	3.00 ± 0.06	295 ± 24	49	409
4	700	710	1.60 ± 0.02	24 ± 2	2.17 ± 0.05	81 ± 7	36	238

  v_g , shear group velocity.

*  The central corneal thickness was measured before and after CXL. The shear group velocity was measured on the ex vivo elastic maps in both treated and untreated areas. The SD corresponds to the spatial heterogeneity of each area. For cases Nos. 1 and 2, the values were deduced from one elastic cross-section (only one imaging plane was acquired). For cases Nos. 3 and 4, the values were deduced from elastic maps of the corneal surface (3-D scans were performed).

Discussion

In this work, we used SSI elastography technique to measure the effect of CXL on the mechanical properties of porcine corneas. 

First, it is important to note that the IOP has a major influence on the corneal elasticity, as shown by prior inflation tests on porcine eyeballs combined with SSI elastography (Nguyen T-M, unpublished data, 2010). The corneal elasticity increases nonlinearly with the IOP. An IOP variation of 3 or 4 mm Hg induces a significant modification of the corneal elasticity that can be detected with SSI elastography. The IOP was thus set to 20 ± 2 mm Hg for all ex vivo experiments. 

We started our study with ex vivo CXL experiments. Only 29% of the treated corneas exhibited significant stiffening (60% ± 20% of the shear wave speed) after CXL. Figure 6 shows a clear separation between the corneas that were affected by the CXL and those that were not affected by this treatment. We suspected the ex vivo conditions would introduce biases in the experiments. In particular, in ex vivo conditions, the stroma does not ensure the water content regulation, as opposed to in vivo conditions. Moreover, the cornea is highly permeable because of the epithelium removal. Since the ultrasound measurements require a total immersion of the sample, a non-negligible amount of water can penetrate the corneal stroma. This phenomenon was confirmed by pachymetric measurements (SP 100, Tomey) performed at the beginning and at the end of the experiments. The initial central corneal thickness (CCT) was 891 ± 152 μm for all the samples. The final CCT was 994 ± 183 μm. Such an edema might accelerate the tissue degradation and disturb the CXL process. The pigs also underwent hot steam at the slaughterhouse, which might result in superficial coagulation on the collected corneas. Nevertheless, some corneas (29%) exhibited significant stiffening after CXL. The distinction between successful and unsuccessful treatments could be due to differences in the postmortem degradation of the tissue properties before the CXL experiments. We did not find any correlation between the initial CCT and the CXL efficiency. But many parameters in our ex vivo protocol might affect the CXL efficiency, such as changes in collagen interfibrillar distance induced by collagen degradation. Thus, to avoid these unknown parameters that clearly affected the efficacy of corneal CXL, we performed in vivo CXL experiments on anesthetized pigs. 

Elastic cross-section maps were acquired in vivo after the treatment, and 3-D elastography scans (acquisitions of different cross-sections) were performed ex vivo on the enucleated eyes. We compared the ex vivo elastic cross-section maps of pairs of eyes coming from the same animal. On each animal, one eye was treated by CXL, while the other one remained untreated. We clearly observed a stiffened area on the treated cornea. Therefore, we demonstrated the ability of the SSI elastography technique to visualize the stiffening effect of CXL. 

Besides, the stiffened area corresponds to the UV-A–beam localization. Outside of the laser focal spot, the elasticity of the treated cornea was similar to that of the untreated cornea. Thus, we showed that the CXL stiffening effect is localized on the UV-A spot. We performed a set of unilateral in vivo CXLs: the epithelium was removed only on the superior half of the cornea. We observed that the UV-A absorption was maximal on the de-epithelized area. This observation is consistent with high-performance liquid chromatography performed by other groups to measure the stromal concentration of riboflavin with and without epithelium removal. ¹¹ This result underlines the importance of the epithelium removal during the CXL procedure. The ex vivo elastic maps obtained after unilateral in vivo CXL exhibited a higher shear wave speed in the treated area than in the untreated area. The geometrical shape and localization of the stiffer area corresponds to the de-epithelized area. 

On the four cases investigated, the shear wave speed was 56% ± 15% (inter-experiments variability) higher in the treated area than in the untreated area (Table). This value is comparable with the ratio obtained in the successful ex vivo CXL experiments. However, the significant variability between the different corneas emphasizes the necessity to monitor the CXL. 

Case No. 4 (Table) is the example shown in this article (Fig. 8). For this case, the shear wave speeds measured ex vivo in treated and untreated regions (2.17 ± 0.05 m/s and 1.60 ± 0.02 m/s, respectively) correspond to the respective stiffness values of 81 ± 7 kPa and 24 ± 1 kPa. These values are deduced from Equation 2, linking the mechanical wave speed and the local stiffness in a guided elastic plate. For these estimations, the thickness was measured on the B-mode images (h = 1 mm ± 50 μm for both areas) and the central frequency (500 Hz) was deduced from the Fourier transform of the displacement field. 

The corneal elasticity measurements reported here are averaged over the entire corneal thickness. Indeed, in our experimental conditions, the radiation force is applied in an area that has an axial length of typically 700 μm up to 1 mm. Thus, we observed an effect of the CXL that is averaged over the entire corneal thickness. 

Experimental improvements must be done to obtain in vivo 3-D scans. Indeed, several imaging planes must be acquired to perform such a scan for a given position of the cornea. In the present study, the anesthesia of the pigs induced uncontrollable motions such as global rotation of the eyeball. From the perspective of clinical applications, a solution to keep the eyeball in a given position consists of asking the patient to focus on a luminous target. 

Previous studies have shown that the viscosity of soft tissues can also be measured using SSI. ^10,12 The corneal viscosity can be deduced from the estimation of the shear wave attenuation. Further studies will be conducted to investigate the feasibility and clinical relevancy of corneal viscosity assessment. 

Conclusions

This study demonstrates the feasibility of using SSI for the in vivo monitoring of CXL. SSI elastography provided, to our knowledge, the first in vivo experimental evidence of the corneal biomechanical changes induced by CXL. CXL was applied in vivo on anesthetized pigs. Elastic maps of the cornea were obtained after treatment in both ex vivo and in vivo conditions. On these maps, the treated area was clearly delineated and exhibited significant stiffening compared with the untreated area (56% ± 15%). Three-dimensional scans provide elastic maps of the corneal surface, enhancing the visualization of the treated area. This technique seems promising for the in vivo monitoring of CXL efficiency as it could be performed immediately after the treatment. Further studies will investigate the usefulness of SSI in performing preoperative diagnosis. Three-dimensional elastic maps of the cornea could help the management of keratoconus or improve the planning of refractive surgery procedures. 

Acknowledgments

This work was supported by the French National Research Agency (ANR MicroElasto). The authors are grateful to the IMM research group (Paris, France; in the public domain at www.imm-recherche.com) for their support and fruitful discussions. The authors would also like to thank the slaughterhouse Etablissements Guy Harang (Paris, France) for providing enucleated porcine eyes. 

Footnotes

 Supported by the French National Research Agency (ANR MicroElasto).

Footnotes

 Disclosure: T.-M. Nguyen, None; J.-F. Aubry, None; D. Touboul, None; M. Fink, None; J.-L. Gennisson, None; J. Bercoff, SuperSonic Imagine (E); M. Tanter, SuperSonic Imagine (C)