# paper.pdf ## Metadata - PDFFormatVersion=1.6 - IsLinearized=true - IsAcroFormPresent=false - IsXFAPresent=false - IsCollectionPresent=false - IsSignaturesPresent=false - CreationDate=D:20260709100839+05'30' - ModDate=D:20260709100915+05'30' - Producer=Acrobat Distiller 10.0.0 (Windows) - pdf:keywords="corneal crosslinking"; "accelerated protocols"; "corneal biomechanics"; "optical coherence elastography"; "wave propagation" - dc:format=application/pdf - dc:title=Air-Coupled Ultrasonic Optical Coherence Elastography Reveals Protocol-Dependent Corneal Stiffening in Epi-On and Epi-Off Riboflavin Crosslinking - dc:identifier=doi:10.1167/tvst.15.7.14 - dc:description=Translational Vision Science and Technology, 2026. doi:10.1167/tvst.15.7.14 - prism:doi=10.1167/tvst.15.7.14 - prism:url=https://doi.org/10.1167/tvst.15.7.14 - prism:aggregationtype=journal - prism:publicationname=Translational Vision Science and Technology - prism:issn=2164-2591 - dc:creator=Fernando Zvietcovich,Alejandra Varea-Bejar,Vlada Pleshcheva,Andrea Curatolo,Judith S. Birkenfeld,Susana Marcos - dc:subject=corneal crosslinking,accelerated protocols,corneal biomechanics,optical coherence elastography,wave propagation - xmpmm:documentid=uuid:405444a2-9c1e-36b4-a854-b7b6dab4d906 - xmpmm:instanceid=uuid:a093cf97-68ab-3fad-a46e-5eb443156c12 ## Contents ### Page 1 Cornea & External DiseaseAir-Coupled Ultrasonic Optical Coherence Elastography Reveals Protocol-Dependent Corneal Stiffening in Epi-On and Epi-Off Riboflavin Crosslinking Fernando Zvietcovich1, Alejandra Varea-Bejar2, Vlada Pleshcheva1, Andrea Curatolo2,3, Judith S. Birkenfeld2, and Susana Marcos2,41 Department of Engineering, Pontificia Universidad Catolica del Peru, Lima, Peru 2 “Daza de Valdes” Institute of Optics, Spanish National Research Council, Madrid, Spain 3 Department of Physics, Politecnico di Milano, Milan, Italy 4 Center for Visual Science, The Institute of Optics, Flaum Eye Institute, University of Rochester, NY, USACorrespondence: Fernando Zvietcovich, Pontificia Universidad Catolica del Peru, Av. Universitaria 1801, San Miguel, Lima 15088, Peru. e-mail: fzvietcovich@pucp.edu.peReceived: October 15, 2025Accepted: May 6, 2026Published: July 10, 2026Keywords: corneal crosslinking; accelerated protocols; corneal biomechanics; optical coherence elastography; wave propagationCitation: Zvietcovich F, Varea-Bejar A, Pleshcheva V, Curatolo A, Birkenfeld JS, Marcos S. Air-coupled ultrasonic optical coherence elastography reveals protocol-dependent corneal stiffening in epi-on and epi-off riboflavin crosslinking. Transl Vis Sci Technol. 2026;15(7):14, https://doi.org/10.1167/tvst.15.7.14Purpose: To evaluate the biomechanical impact of corneal crosslinking (CXL) when combining three different ultraviolet A (UV-A) riboflavin (RB) photosensitizers and proto- cols (Dresden protocol [DP] and accelerated protocol [AP]) using air-coupled ultrasonic optical coherence elastography in ex vivo rabbit corneas.Methods: An air-coupled ultrasound excitation optical coherence elastography system was used to excite the corneal apex and generate Lamb wave propagation along 16 corneal cross-sectional meridians. Measurements were conducted on ex vivo rabbit eyes (n = 45) during three treatment phases: Virgin, after 30 minutes of RB soaking, and after UV-A irradiation. The protocols used include the DP (3 mW/cm2, 30 minutes) and two accelerated protocols (AP-1: 9 mW/cm2, 10 minutes; and AP-2: 30 mW/cm2, 3 minutes). Each protocol was tested with three photosensitizers: epithelium-on (TE), and epithelium-off with hyperosmolar (M) and hypo-osmolar solutions (D). Lamb wave speed and average corneal thickness were calculated for each meridian to estimate meridian-dependent corneal shear modulus. Differences were statistically analyzed using linear mixed-effects regression.Results: The DP-D produced the strongest significant corneal stiffness increase (480.5 kPa; P < 0.001) during the UV irradiation phase, and the most pronounced corneal thinning (186 μm; P < 0.001). DP-TE achieved the next strongest stiffening (309.5 kPa) during UV irradiation, with no significant thickness change. All other protocol– photosensitizer combinations did not achieve significant corneal stiffening except for AP-1-D and AP-2-D during the RB soaking phase.Conclusions: Accelerated protocols did not produce significant corneal stiffening for any RB photosensitizer except when using dextran-based D. DP-D and DP-TE produced the greatest balance between shear modulus increase and corneal thickness decrease, suggesting DP-TE as a potential compromise between biomechanical impact, corneal integrity, and faster postoperative recovery.Translational Relevance: This work quantified the impact of clinically used crosslinking photosensitizers and protocols on corneal stiffening using an air-coupled ultrasound excitation optical coherence elastography, which holds promise for clinical assessments in patients. Copyright 2026 The Authors tvst.arvojournals.org | ISSN: 2164-2591 1 This work is licensed under a Creative Commons Attribution 4.0 International License. ### Page 2 Corneal Stiffening in Accelerated UV-A Crosslinking TVST | July 2026 | Vol. 15 | No. 7 | Article 14 | 2IntroductionCorneal crosslinking (CXL) has emerged as a criti- cal therapeutic intervention in ophthalmology, partic- ularly for treating corneal ectasia and keratoconus.1 By inducing photochemical CXL of stromal collagen, CXL enhances corneal rigidity and stabilizes its shape, preventing the progression of biomechanical deterio- ration associated with these pathologies.2 The origi- nal Dresden protocol (DP), which involves ultravi- olet A (UV-A) irradiation combined with riboflavin (RB) as a photosensitizer, remains the gold standard for CXL treatments.3 However, there are concerns regarding the duration of the procedure (1 hour) and the induced corneal thinning that may induce patient discomfort and risk of corneal damage. Alter- natives to overcome this problem include the use of different irradiation wavelengths and photosensitiz- ers (e.g., green light and Rose Bengal),4 or the same UV-A with different irradiation regimes (accelerated protocols [APs]) and osmolarity of the photosensitizer solution. UV-A–accelerated CXL protocols and new RB preparations seek to optimize treatment by reduc- ing the total procedure time (3–10 minutes), avoiding corneal thickness reduction while preserving biome- chanical efficacy.5 Besides, removing the epithelium in corneal CXL enhances RB penetration and UV- A efficacy, increasing treatment effectiveness; however, it also increases the risk of postoperative pain, infec- tion, and delayed epithelial healing. For this reason, protocols aimed at preserving the epithelium have been proposed.6 The biomechanical effects of CXL are primar- ily characterized by increased corneal stiffness and enhanced resistance to enzymatic degradation. CXL induces structural alterations within the collagen fibril network, strengthening interfibrillar bonds and thereby improving the overall mechanical integrity of the cornea.2,5,7 Biomechanically, the corneal stroma can be approximated as a transversely isotropic material,8,9 whereby CXL-induced stiffening manifests as an increase in both the out-of-plane shear modulus (relative to corneal surface), and the in-plane Young’s modulus (parallel to corneal surface).8 While the conventional DP produces a marked and uniform stiff- ening effect,10–13 APs tend to result in a more superficial CXL, raising concerns about their capacity to achieve equivalent corneal biomechanical reinforcement.5,14–18 To address the limitations of the DP, APs were developed to shorten treatment time by increasing the UV-A irradiance while maintaining, a constant total energy dose of 5.4 J/cm2 across all cases. Although standard CXL uses an irradiance of 3 mW/cm2 for 30 minutes,3 APs use higher intensities (9–30 mW/cm2) with shorter exposure times (3–10 minutes).5 These accelerated variations are based on the Bunsen–Roscoe law of reciprocity, which states that the photochemi- cal effect of light exposure is directly proportional to the total energy received, independent of duration.19 However, this reciprocity does not hold linearly in corneal CXL, as oxygen diffusion becomes limited at higher irradiances, resulting in a reduced CXL effect.20,21 Parallel to procedural advances, considerable effort has been directed toward developing novel RB formu- lations to modulate corneal hydration and enhance stromal penetration during CXL. Traditional dextran- based solutions, such as those used in the original DP, facilitate controlled corneal hydration to optimize CXL depth, but they are associated with significant intraoperative corneal thinning.5 In contrast, emerg- ing dextran-free formulations have been designed to enhance RB diffusion into the stroma while minimiz- ing dehydration, thereby supporting the feasibil- ity of transepithelial (epi-on) CXL approaches.22–24 However, the interaction between these formulations and specific CXL protocols may substantially influ- ence treatment efficacy, with some studies indicat- ing reduced biomechanical reinforcement relative to conventional epithelium-off (epi-off) protocols.14–17,25 Therefore, the quantitative evaluation of stromal stiff- ening across novel RB formulations and protocols, benchmarked against the Dresden standard, is criti- cal for elucidating the trade-off between biomechanical strengthening and preservation of corneal integrity— particularly under conditions approximating clinical practice. This study presents a comparative analysis of the biomechanical impact of UV-A RB-mediated corneal CXL in ex vivo rabbit corneas maintained at a physiological intraocular pressure (IOP) of 15 mm Hg. Three irradiation protocols were evalu- ated: the standard DP (30 minutes, 3 mW/cm2) and two APs (10 minutes of irradiation, 9 mW/cm2; and 3 minutes of irradiation, 30 mW/cm2), each combined with one of three RB photosensitizers: epithelium-off using either a dextran-based or dextran-free formu- lation, and epithelium-on (TE) using a transepithe- lial dextran-free solution. Biomechanical parameters were assessed at three stages: baseline (Virgin), post- RB soaking (Post-RB), and post-irradiation (Post- CXL). Using air-coupled ultrasound excitation optical coherence elastography (ACUS-OCE),26–28 we quanti- fied changes in the corneal shear modulus and thick- ness changes across protocol–photosensitizer combi- nations. These measurements aimed to elucidate the balance between mechanical reinforcement and struc- tural preservation, informing optimization strategies for future clinical translation of CXL techniques. ### Page 3 Corneal Stiffening in Accelerated UV-A Crosslinking TVST | July 2026 | Vol. 15 | No. 7 | Article 14 | 3Material and Methods Sample PreparationForty-five freshly enucleated rabbit eyes from adult New Zealand white rabbits (2–3 kg) were obtained (Facultad de Veterinaria, Complutense University of Madrid, Madrid, Spain) and used for experimental procedure within 24 hours post mortem. The anatom- ical orientation was identified during the eye enucle- ation process and marked for subsequent location into the eye holder. Before measurements, muscles and conjunctival tissue were removed from the sclera. After preparation, the intact eye was placed into a customized holder that allows for ocular canulation using two 25G needles that penetrated into the anterior chamber through the limbus along nasal and temporal regions (Fig. 1a). The eyes were systematically aligned so that their anatomical superior–inferior meridian coincides with the vertical scanning axis in the holder. One needle was connected to a pressure sensor, and the other was connected to a controlled fluid pump system. Both the pressure sensor and the microinfusion pump were fully controlled by the computer through a closed loop system that ensured constant IOP of 15 ±0.2 mm Hg during all measurements.26Corneal CXL Treatments and ProtocolsBiomechanical outcomes were investigated for three CXL protocols (DP and two APs, AP-1 and AP- 2) each combined with three RB photosensitizers:Figure 1. (a) Pressurized rabbit ex vivo eye globe using a closed loop IOP control system set to 15.0 ± 0.2 mm Hg. (b) ACUS-OCE system setup composed of a PhS-SS-OCT system and a 500-kHz ACUS transducer, confocal and coaligned with the OCT sample arm. The acoustic excitation is focused on the corneal apex. (c) Acquisition of 8 OCT meridians distributed along 360° (steps of 22.5°), forming 16 semimeridians centered at the corneal apex. (d) Outward antisymmetric 0-mode Lamb wave propagation along a single corneal meridian after excitation. I, inferior; N, nasal; S, superior; T, temporal. ### Page 4 Corneal Stiffening in Accelerated UV-A Crosslinking TVST | July 2026 | Vol. 15 | No. 7 | Article 14 | 4Table 1. Summary of Corneal CXL Photosensitizers Photosensitizers Transepithelial Dextran RB Protocol Used D No 20% 0.1% DP, AP-1, AP-2 M No 0% 0.1% DP, AP-1, AP-2 TE Yes 0% 0.25% DP, AP-1, AP-2Table 2. Summary of Corneal CXL Protocols Protocol RB Soaking, Minutes UV-A Irradiation, Minutes Irradiance, mW/cm2RB Application During UV-A Photosensitizers Used DP 30 30 3 Every 5 min D, M, TE AP-1 30 10 9 Every 3 min D, M, TE AP-2 30 3 30 Every 1 min D, M, TE(D) a dextran-based, epi-off solution (0.1% RB, 20% dextran: Dresden solution, isotonic; MedioCross Glaukos Corp, Aliso Viejo, CA); (M) a dextran- free epithelium-off solution (epi-off, 0.1% RB, 1.1% hydroxypropyl methylcellulose [HPMC], isotonic); and (TE) transepithelial dextran-free solution for TE appli- cation (epi-on, 0.25% RB, 1.2% HPMC, 0.01% benza- lkonium chloride [BKC]), as summarized in Table 1. The protocols used in this study included the DP and two accelerated (AP-1, AP-2) CXL protocols, maintaining a constant total energy dose of 5.4 J/cm2 in all cases, as described in Table 2. A total of 45 eyes were distributed in different CXL sessions obtained from the combination of all protocols (DP, AP-1, and AP-2), and all RB photosensitizers (D, M, and TE), resulting in 5 eyes in each of the 9 [protocol]–[photosensitizer] cases. UV-A irradiation was delivered using a commercial LED (365 nm central wavelength, 9 nm full-width half- maximum bandwidth) lamp mounted with an aspheric collimation lens (EFL = 40 mm, AR-coated) compat- ible with Leica microscopes (M365LP1-C2, Thorlabs, Newton, NJ) yielding a 37-mm beam diameter and a total power of 435 mW. A custom-designed three- dimensionally printed mask was positioned anterior to the light source to confine the effective irradiation area to a 15-mm diameter circle and exposure of the limbus. The lamp was operated in a continuous-wave mode and calibrated with a UV radiometer (UVA-365SD, Lutron Electronic, Coopersburg, PA) to provide irradiances of 3 mW/cm2, 9 mW/cm2, and 30 mW/cm2, as specified by each CXL protocol.ACUS-OCE SetupBiomechanical measurements of the corneas were obtained using a phase-sensitive swept-source optical coherence tomography (PhS-SS-OCT) system, coaligned in a confocal configuration with a custom- built air-coupled ultrasonic (ACUS) piezoelectric transducer for tissue excitation at the corneal apex,29–31 as shown in Figure 1b. The PhS-SS-OCT is a custom- built system described in detail in previous publica- tions32–34 featuring a swept source (VCSEL, SL132120, Thorlabs) centered at a 1300-nm wavelength, with a spectral bandwidth of 50 nm, and an A-line scan rate of 200 kHz. The full-width half-maximum axial resolution is 16 μm in air, and a lateral resolution (sample arm) of 40 μm at the focal plane, and a 5.15-mm imaging depth of field in tissue.32 Tissue nanometer displacement was obtained in the sample over time in response to a stimulation using the phase information of the OCT spectrum through the phase-sensitive mode.35 The system allows multi- directional scanning via high-speed galvanometric scanning mirrors (Saturn 1B, Scanner MAX, Pangolin, New York, NY). In this study we acquired 8 corneal cross-sectional meridians (10 mm length) equally distributed over 360° (steps of 22.5°), forming 16 semimeridians centered at the corneal apex, as shown in Figure 1c. The ultrasonic transducer was driven by a three- cycle tone burst, consisting of a 2.5-kHz train of square pulses modulating a 500-kHz carrier signal that was preamplified using a radiofrequency ampli- fier (100A250A, Amplifier Research, Souderton, PA). The excitation generated a radially symmetric acous- tic radiation force at the corneal apex, producing a localized push with a lateral extent of 0.6 mm. The resulting deformation (∼1 μm) in the cornea initi- ated the outward propagation of a cylindrical Lamb wave, specifically the antisymmetric zero-order mode (A0), as illustrated in Figure 1d. Lamb wave propa- gation, along with structural imaging of the cornea, was acquired along each semimeridian using the PhS- SS-OCT system. The 15-mm central aperture of the ### Page 5 Corneal Stiffening in Accelerated UV-A Crosslinking TVST | July 2026 | Vol. 15 | No. 7 | Article 14 | 5Figure 2. OCE processing approach. (a) Motion (particle velocity) wave propagation videos of each meridian. A representative frame at t0= 0.5 ms is shown for each video. (b) Structural B-mode OCT images of the cornea of each meridian. We defined a left- and right-propagation spatial regions (4 mm × 4 mm size) in each B-mode image as shown om the green rectangular region of (b). Average corneal thickness polar plot (in μm) extracted from the left- and right-propagation spatial regions of (b) using thresholding techniques. (d) Space–time motion maps depicting wave propagation extracted form (a) at the corneal surface. Left- and right-propagation spatiotemporal regions (4 mm × 1.75 ms size) were defined (discontinuous-line boxes). (e) Lamb wave phase speed estimation vs. frequency obtained from the two-dimensional Fourier analysis on each spatiotemporal propagating region in (d). Phase speeds at 2.5 kHz were extracted for each propagation side. (f) Lamb wave speed cLpolar plot (in meters per second) was obtained by analyzing (e) in each corneal meridian. (g) Shear modulus (G) polar plot (in kPa) was calculated from (f) and (c) using the mRLFE model.36ACUS transducer permitted unobstructed passage of the OCT beam, enabling co-localized optical access to the corneal sample.OCE Measurements and ProcessingThe M-B scanning protocol35 was used to capture wave propagation along eight cross-meridians of the cornea. At each lateral location, the ACUS was triggered and an M-mode frame consisting of 450 A- lines over time (temporal sampling interval t = 5 μs) was acquired. This M-mode acquisition was repeated for each of the 100 spatial positions forming a 10-mm a cross-meridian (spatial resolution = 0.1 mm) and subsequently repeated for all eight cross-meridians. A 250-μs interval was implemented between consecutive steps to allow for stable lateral and angular transi- tions across all scanning positions (Fig. 1c). The total acquisition time was 2 seconds, including the transition periods allocated to the galvanometric scanning system to move smoothly between meridians, thereby minimiz- ing mechanical oscillations and preventing mirror- induced vibrations. Each cornea was measured using the ACUS-OCE system at three time points along the CXL treatment, also called treatment phases: (1) Virgin (referring to untreated epi-off cornea for D and M; and epi-on cornea for TE); (2) Post-RB (after 30 minutes of RB soaking); and (3) Post-CXL (after the UV irradiation). Drops of balanced salt solution were poured over the corneas to hydrate them before each OCE measure- ment. The processing scheme applied to the OCT-based data along each corneal semi-meridian is presented in Figure 2 and follows these steps. (1) From the M-B scanning data, particle velocity (motion) videos ### Page 6 Corneal Stiffening in Accelerated UV-A Crosslinking TVST | July 2026 | Vol. 15 | No. 7 | Article 14 | 6(Fig. 2a) were reconstructed for each of the eight meridians, along with corresponding B-mode struc- tural OCT images (Fig. 2b). Each motion video shows wave propagation toward the left and right direction as the excitation was performed in the corneal apex; therefore, we defined a left- and right-propagation spatial regions (4 mm × 4 mm size) in each B-mode image as shown green rectangular region of Figure 2b. (2) The average corneal thickness was computed in the left- and right-propagation spatial regions of each meridian by applying threshold-based segmentation to the B-mode images, enabling the generation of a polar map of the average corneal thickness (in micrometers) (Fig. 2c). (3) Space–time maps were constructed by tracking A0 Lamb waves from the motion videos along each meridian (Fig. 2d). Left- and right-propagation spatiotemporal regions (covering 4-mm in space and 1.75 ms in time) were defined (discontinuous line boxes in Fig. 2d). (4) A two- dimensional Fourier transform was applied separately to the left- and right-propagation spatiotemporal regions of the space–time map to estimate the frequency-dependent phase speed within each region35 (Fig. 2e). (5) The Lamb wave phase speed at 2500 Hz was extracted from the frequency-domain speed plot from both left- and right-propagation waves, and the resulting values were used to generate a polar plot of phase speed (in meters per second) across all eight meridians (Fig. 2f).Shear Modulus EstimationThe shear modulus was estimated using the modified Rayleigh–Lamb frequency equation (mRLFE) model,36,37 approximating the cornea as an isotropic, homogenous, and viscoelastic material. Although the cornea is inherently anisotropic and consists of layered microstructures,9,38 this first- order approximation facilitates direct comparisons of the biomechanical effects of various CXL treat- ments and protocols under consistent assumptions. For each cornea, the mRLFE mode was applied, using 16 paired measurements of Lamb wave phase speed and local corneal thickness obtained along 16 semimeridians. The model parameters included an excitation frequency of 2500 Hz, Poisson’s ratio of 0.498 (assumed nearly incompressible), corneal tissue density of 988 kg/m3, aqueous humor density of 1000 kg/m3, and pressure-wave speed of 1500 m/s in cornea and aqueous humor. This analysis yielded 16 shear modulus estimates (in kilopascals), one for each semimeridian, visualized as a polar plot in Figure 2g.Statistical AnalysisChanges in corneal thickness and shear modulus, were analyzed using linear mixed-effects regression (LME)39 models, which account for the hierarchi- cal and repeated-measures structure of the data. The models included fixed effects (protocol, photosen- sitizer, treatment phase, and angular position) and random effects (repeated measurements within each cornea) to capture subject-specific variability. This statistical framework enables robust inference based on analysis of variance (ANOVA) P values, and confi- dence intervals (CIs), providing an accurate and biolog- ically consistent assessment of corneal biomechanical responses across experimental conditions. The general LME model formulation is given by:Yi jkl = μ + βP · Pi + βL · L j+ βT · Tk + βA · Am+ βPL · (P × L)i j + βPT · (P × T )ik+ βLT · (L × T ) jk + βPLT · (P × L × T )i jk+ βPA · (P × A)im + Rl + bl Pi + i jkl (1) where Yijkl is the response variable (shear modulus or thickness); μ is the overall intercept (grand mean); Pi is the fixed effect of treatment phase (Virgin, Post-RB, or Post-CXL) with βP as the estimate coefficient; Lj is the fixed effect of protocol (DP, AC-1, or AC-2) with βLas the estimate coefficient; Tk is the fixed effect of RB photosensitizer (D, M, or TE) with βT as the estimate coefficient; Am is the fixed effect of an angle (16 merid- ians for one subject); (P × L)ij captures the interac- tion between treatment phase and protocol; (P × T)ikcaptures the interaction between treatment phase and RB photosensitizer; (L × T)jk captures the interaction between protocol and RB photosensitizer; (P × A)imcaptures the interaction between treatment phase and angle; (P × L × T)ijk captures the three-way interaction of treatment phase, protocol and RB photosensitizer;Rl is the random intercept of the cornea, accounting for differences in baseline shear modulus and thickness across individual; blPi is the random slope of treatment phase, allowing individual corneas to exhibit different responses in each treatment phase; and εijkl is the resid- ual error representing unexplained variability in the response variable that is not captured by the fixed or random effects (assumed to be normally distributed). The LME model was fitted to the shear modulus shear modulus and corneal thickness data to gener- ate predicted values (Yijkl) as a function of the fixed effects: protocol, RB photosensitizer and treatment phase. For shear modulus and corneal thickness, 45 eyes × 3 measurement phases × 16 semimeridians ### Page 7 Corneal Stiffening in Accelerated UV-A Crosslinking TVST | July 2026 | Vol. 15 | No. 7 | Article 14 | 7= 2160 data points in each case were used for the LME fitting. To evaluate the performance of the LME model, we computed the marginal R2 (Rm2) and condi- tional R2 (Rc2) values. Rm2 represents the proportion of variance of predictions (Yijkl) explained by the fixed effects alone, and Rc2 includes both fixed and random effects. The type III ANOVA test using Satterthwaite’s method was used to assess the contribution of each fixed effect into shear modulus and thickness predic- tions.Effectiveness Factor (EF) for Corneal CXLThe EF was introduced to determine whether changes in shear modulus scale proportionally with corneal thickness:EFVirgin → Post−CXL = GVirgin → Post−CXLTHVirgin → Post−CXL− 1 (2) where:GVirgin → Post−CXL= GBL + (GPost−CXL − GVirgin)GBL (3) quantifies the relative increase in corneal shear modulus (G) following CXL with respect to the baseline corneal shear modulus GBL calculated as the mean shear modulus of all corneas in the Virgin phase. Values of GVirgin → Post−CXL > 1 indicate a desirable increase in corneal stiffness, while GVirgin → Post−CXL< 1 represents a softening effect of the cornea. Additionally,THVirgin → Post−CXL= THBL + ∣ ∣THPost−CXL − THVirgin∣ ∣THBL(4) accounts for the relative undesired effect of the CXL, reflecting corneal thinning that could pose a risk of UV-damaging corneal endothelium or thickening due to corneal swealing, and this effect could be even more exacerbated by the ex vivo nature of the procedure. Such a change in corneal thickness was normalized with respect to the baseline corneal thickness (THBL) calculated as the mean thickness of all corneas in the Virgin phase. THVirgin → Post−CXL > 1 represents increase or decrease of corneal thickness after CXL. Therefore, by evaluating Equation 1 using Equations 2 and 3, three scenarios can be distin- guished: EF > 0 indicates that the increase in shear modulus exceeds the corresponding change in thick- ness (a desirable CXL effect); EF < 0 indicates that the increase in shear modulus (or possible decrease) is smaller than the thickness change (a undesirable trade-off among CXL effects); and EF = 0 denotes proportional changes in shear modulus and thickness (a neutral effect).Results Dresden ProtocolFigure 3 presents meridian-dependent polar plots of the average (across corneal samples) Lamb wave speed, corneal thickness, and shear modulus shear modulus of corneas in the Virgin, Post-RB, and Post-CXL phases of the DP when using photosensitizers D (Figs. 3a– c), M (Figs. 3d–f), and TE (Figs. 3g–i). Although the distribution of the polar plots across meridians is not necessarily uniform within single treatment phase, changes in Lamb wave speed, shear modulus (G) and average corneal thickness between phases are signifi- cantly uniform across meridians (P > 0.05), except for the cases of Lamb wave speed change (Post-RB–Virgin;P < 0.05), shear modulus change (Post-RB–Virgin; P< 0.01) and shear modulus change (post-CXL − Post- RB; P < 0.05) when using D photosensitizers. Figure 4a presents the mean values of shear modulus (G) across eyes for the three treatment phases and for each RB photosensitizer, with 95% confi- dence limits. The D photosensitizer shows the most pronounced significant increase of shear modulus during the Virgin–Post-CXL phase compared with M and TE. The shear modulus increased by 1054.9 kPa for D (P < 0.001), and 344.6 kPa for TE (P < 0.01), but no significant shear modulus change for the M photosensitizer, as shown in Figure 4b. Moreover, shear modulus increased significantly (574.4 kPa; P < 0.001) after RB soaking (i.e., the change from Virgin to Post- RB), but only for the D photosensitizer, highlight- ing the intrinsic stiffening effect of dextran-based RB solutions even in the absence of UV-A irradiation. Finally, the effect of UV-A irradiation on the increase in shear modulus (i.e., the change from Post-RB to Post-CXL) was significant for D photosensitizer (G =480.5 kPa; P < 0.001) and the TE photosensitizer (G= 309.5 kPa; P < 0.05), while no significant change was observed for the M treatment. Figure 4c shows the average tendency of corneal thickness across treatment phases for each RB photo- sensitizer. The D photosensitizer show the greatest thinning in corneal thickness from the Virgin to Post- CXL phase, with a reduction of 186 μm (P < 0.001), while M produced a significant change of 38.7 μm (P <0.05) only in the Virgin to Post-RB phase, and no signif- icant change in TE (Fig. 4d). Across all treatments, ### Page 8 Corneal Stiffening in Accelerated UV-A Crosslinking TVST | July 2026 | Vol. 15 | No. 7 | Article 14 | 8Figure 3. DP multimeridian estimations. Group average meridian-dependent polar plots of Lamb wave speed (left), average corneal thick- ness (center), and estimated shear modulus (right) are shown in corneas in the Virgin, Post-RB, and Post-CXL phases when using photosen- sitizers D (a, b, c), M (d, e, f), and TE (g, h, i) and DP (3 mW/cm2, 30 minutes of UV). Scattered points shown mean values within a group. Standard errors (SD) for each meridian are shown in the plots as shadow patterns and bars. ### Page 9 Corneal Stiffening in Accelerated UV-A Crosslinking TVST | July 2026 | Vol. 15 | No. 7 | Article 14 | 9Figure 4. DP statistical results. (a) Estimated average shear modulus with 95% confidence limits vs. treatment phase for all RB photosensi- tizers. (b) LME-predicted mean difference of shear modulus across phases: Virgin → Post-RB, Post-RB → Post-CXL, and Virgin → Post-CXL for all RB photosensitizers. (c) Estimated average corneal thickness with 95% confidence limits vs. treatment phase for all RB photosensitizers. (d) LME-predicted mean difference of corneal thickness across phases: Virgin → Post-RB, Post-RB → Post-CXL, and Virgin → Post-CXL for all RB photosensitizers. Significant differences are presented as ***P < 0.001; **P < 0.01; and *P < 0.05. ns, not significant.most thinning occurred during the Virgin–Post-RB phase. In every case, corneal thickness changes between phases along the corneal meridians were spatially uniform (P > 0.05), indicating that the CXL-induced thinning was uniform across the corneal surface.Accelerated Protocol 1Figure 5 presents meridian-dependent polar plots of the average (across corneal samples) Lamb wave speed, corneal thickness, and shear modulus shear modulus of corneas in the Virgin, Post-RB, and Post-CXL phases of the AP-1 when using photosensitizers D (Figs. 5a–c), M (Figs. 5d–f), and TE (Figs. 5g–i). Similar to the DP protocol, changes in Lamb wave speed, shear modulus (G), and average corneal thickness between phases are significantly uniform across meridians (P > 0.05), except for the case of Lamb wave speed change (Post- CXL−Virgin; P < 0.05), when using the M photosen- sitizer. Figure 6a presents the mean values of shear modulus (G) across corneas for the three treatment phases and for each RB photosensitizer, with 95% confidence limits. In the AP-1 protocol, only the D formulation produced a significant increase of shear modulus (336.4 kPa; P < 0.001) from the Virgin to the ### Page 10 Corneal Stiffening in Accelerated UV-A Crosslinking TVST | July 2026 | Vol. 15 | No. 7 | Article 14 | 10Figure 5. AP-1 multimeridian estimations. Group average meridian-dependent polar plots of Lamb wave speed (left), average corneal thickness (center), and estimated shear modulus (right) are shown in corneas in the Virgin, Post-RB, and Post-CXL phases when using photo- sensitizers D (a, b, c), M (d, e, f), and TE (g, h, i) and AP-1 (9 mW/cm2, 10 minutes of UVA). Scattered points shown mean values within a group. Standard errors (SD) for each meridian are shown in the plots as shadow patterns and bars. ### Page 11 Corneal Stiffening in Accelerated UV-A Crosslinking TVST | July 2026 | Vol. 15 | No. 7 | Article 14 | 11Figure 6. AP-1 statistical results. (a) Estimated average shear modulus with 95% confidence limits vs. treatment phase for all RB photosen- sitizers. (b) LME-predicted mean difference of shear modulus across phases: Virgin → Post-RB, Post-RB → Post-CXL, and Virgin → Post-CXL for all RB photosensitizers. (c) Estimated average corneal thickness with 95% confidence limits vs. treatment phase for all RB photosensitizers. (d) LME-predicted mean difference of corneal thickness across phases: Virgin → Post-RB, Post-RB → Post-CXL, and Virgin → Post-CXL for all RB photosensitizers. Significant differences are presented as ***P < 0.001; **P < 0.01; and *P < 0.05. ns, not significant.Post-CXL phase change, whereas M and TE showed no significant shear modulus change in any phase (Fig. 6b). Furthermore, with the D photosensitizer, the only significant increase in shear modulus (162.1 kPa; P< 0.01) occurred during the RB soaking phase (Virgin– Post-RB), with no additional stiffening effect after UV- A irradiation (Post-RB–Post-CXL). Figure 6c shows the average tendency of corneal thickness across treatment phases for each RB photo- sensitizer. Significant corneal thinning was only observed in the D treatment, with reductions of 88.8 μm (P < 0.001) between the Virgin–Post-RB and of 94.7 μm (P < 0.001) between Virgin and Post-CXL. In contrast, the M treatment (epi-off) produced moder- ate corneal thickness increase of 51.2 μm (P < 0.05) between the Virgin and the Post-CXL phase, possibly due to the corneal swealing due to hydration during the experiments (Fig. 6d). No significant corneal thickness changes were observed for the TE photosensitizer in any treatment phase.Accelerated Protocol 2Figure 7 presents meridian-dependent polar plots of the average (across corneal samples) Lamb wave speed, corneal thickness, and shear modulus of corneas ### Page 12 Corneal Stiffening in Accelerated UV-A Crosslinking TVST | July 2026 | Vol. 15 | No. 7 | Article 14 | 12Figure 7. AP-2 multimeridian estimations. Group average meridian-dependent polar plots of Lamb wave speed (left), average corneal thickness (center), and estimated shear modulus (right) are shown in corneas in the Virgin, Post-RB, and Post-CXL phases when using photo- sensitizers D (a, b, c), M (d, e, f), and TE (g, h, i) and AP-2 (3 mW/cm2, 3 minutes of UVA). Scattered points shown mean values within a group. Standard errors (SD) for each meridian are shown in the plots as shadow patterns and bars. ### Page 13 Corneal Stiffening in Accelerated UV-A Crosslinking TVST | July 2026 | Vol. 15 | No. 7 | Article 14 | 13Figure 8. AP-2 statistical results. (a) Estimated average shear modulus with 95% confidence limits vs. treatment phase for all RB photosen- sitizers. (b) LME-predicted mean difference of shear modulus across phases: Virgin → Post-RB, Post-RB → Post-CXL, and Virgin → Post-CXL for all RB photosensitizers. (c) Estimated average corneal thickness with 95% confidence limits vs. treatment phase for all RB photosensitizers. (d) LME-predicted mean difference of corneal thickness across phases: Virgin → Post-RB, Post-RB → Post-CXL, and Virgin → Post-CXL for all RB photosensitizers. Significant differences are presented as ***P < 0.001; **P < 0.01; and *P < 0.05. ns, not significant.in the Virgin, Post-RB, and Post-CXL phases of the AP-2 when using photosensitizers D (Figs. 7a–c), M (Figs. 7d–f), and TE (Figs. 7g–i). Changes in Lamb wave speed, shear modulus (G), and average corneal thickness between phases are significantly uniform across meridians (P > 0.05), except for the D case in Lamb wave speed and shear modulus changes along all phases (P < 0.001 for all cases). Figures 7a and 7c show a referable stiffening direction along 60° and 90°. Figure 8a shows the mean trends of shear modulus across treatment phases for each photosensitizer. In the AP-2 protocol, a significant increase in shear modulus (750.4 kPa; P < 0.001) was observed also during RB soaking (Virgin–Post-RB) with the D photosensitizer; no significant changes were detected for M and TE (Fig. 8b). Interestingly, for the D formulation, shear modulus decrease significantly during the irradiation phase (−307.4 kPa; P < 0.05), offsetting the previ- ous gain and resulting in a net stiffening of 443 kPa (P < 0.001) by the end of the treatment (Virgin–Post- CXL). Figure 8c presents the trends in corneal thick- ness across treatment phases for each photosensitizer formulation. Significant corneal thinning was observed only with the D photosensitizer, showing reductions of 195.2 μm (P < 0.001) between the Virgin and the Post-RB phases, and 161.9 μm (P < 0.001) between ### Page 14 Corneal Stiffening in Accelerated UV-A Crosslinking TVST | July 2026 | Vol. 15 | No. 7 | Article 14 | 14Table 3. Type III ANOVA (Satterthwaite Method) for Fixed Effects in the LME Model of Shear Modulus and Corneal Thickness Shear Modulus Corneal Thickness Fixed Effect F-Statistic P Value Ranking F-Statistic P Value Ranking Phase 46.81 <0.001 1 51.66 <0.001 2 Phase × photosensitizer 46.45 <0.001 2 55.17 <0.001 1 Photosensitizer 40.63 <0.001 3 23.25 <0.001 3 Angle 19.73 <0.001 4 3.95 <0.001 6 Protocol 13.98 <0.001 5 2.53 ns Phase × protocol 11.6 <0.001 6 8 <0.001 4 Phase × protocol × photosensitizer 6.31 <0.001 7 4.06 <0.001 5 Protocol × photosensitizer 5.84 <0.001 8 2.05 ns Phase × angle 3.15 <0.001 9 0.62 ns ns: non-significant.the Virgin and Post-CXL phases. In contrast, the M formulation (epi-off) exhibited a moderate increase in thickness (52.1 μm; P < 0.05) from Virgin to Post-CXL, likely due to corneal hydration modulation during the experiments (see Fig. 8d). No significant corneal thick- ness changes were observed for the TE photosensitizer in any treatment phase.LME Regression ModelThe LME model for shear modulus yielded a marginal R2 (R2m) of 0.50 and a conditional R2 (R2c) of 0.68, indicating that fixed effects explain half of the variability in shear modulus, while subject- specific random effects provide a meaningful additional contribution. In contrast, corneal thickness showed a marginal R2 of 0.60 and a markedly higher conditionalR2 of 0.95, demonstrating that interindividual variabil- ity dominates thickness responses beyond the effects of experimental factors. The ANOVA results show that, for shear modulus (see Table 3), all main effects and interactions are signif- icant. The three most ranked effects are phase (wp= 46.80; P < 0.001); phase × photosensitizer (F =46.45; P < 0.001); and photosensitizer (F = 40.63; P< 0.001). Angular position also has a strong effect (F= 19.73; P < 0.001), indicating that shear modulus varies significantly across different semimeridians of the cornea as shown by the nonuniform distribu- tions in Figures 3c, 3f, and 3i; Figures 5c, 5f, and 5i; and Figures 7c, 7f, and 7i. All interaction terms are also significant, including phase × protocol (F = 11.60;P < 0.001), protocol × photosensitizer (F = 5.84; P< 0.001), and the three-way interaction (F = 6.31; P< 0.001). Notably, the phase × angle interaction is significant (F = 3.15; P < 0.001), indicating that some spatial pattern of shear modulus evolves across treat- ment phases as reported in the cases of Figures 3a and 3c; Figure 5d; and Figures 7a and 7c. In contrast, corneal thickness is primarily driven by phase × photosensitizer (F = 55.17; P < 0.001), phase (F = 51.66; P < 0.001), and photosensitizer (F = 23.25;P < 0.001), as shown in Table 3. Protocol alone is not significant (F = 2.52; P = 0.092); however, the signifi- cant phase × protocol interaction (F = 8.00; P < 0.001) shows that protocol-dependent differences emerge over the course of treatment. Although thickness may vary across angular positions (angular effect, F = 3.95; P< 0.001), the phase × angle interaction is not signifi- cant (F = 0.62; P = 0.947), indicating that this spatial pattern remains stable across treatment phases.Effectiveness FactorThe EF was computed using LME-predicted values for all protocol–photosensitizer combinations without restricting the analysis to statistically significant effects, to preserve comparability across conditions. Because the EF is derived from estimated changes in shear modulus and corneal thickness, its reliability depends on the precision of these underlying model estimates. Consequently, EF values associated with nonsignifi- cant contrasts or wide CIs should be interpreted with caution, because they may reflect variability rather than robust biomechanical effects. To quantify this uncertainty, CIs for EF were obtained via bootstrap resampling (5000 iterations) of the LME-predicted means for shear modulus and thickness. At each iteration, values were sampled from the estimated distributions defined by the model-based means and standard errors, and EF was recomputed using the formulation in Equations 2, 3, and 4. This ### Page 15 Corneal Stiffening in Accelerated UV-A Crosslinking TVST | July 2026 | Vol. 15 | No. 7 | Article 14 | 15Table 4. EF Calculated for all Protocol Photosensitizer Cases (Equation 2) by Computing the Change of Shear Modulus (G), and Change of Thickness (TH) as Indicated in Equations 3 and 4, Respectively Protocol- Photosensitizer GVirgin, kPa GPost-CXL, kPa THVirgin, μm THPost-CXL, μm EF Lower 95% CI of EF Upper 95% CI of EF DP-D 166.654 1221.613 499.537 313.534 4.208 3.051 5.485 DP-TE 258.124 602.765 459.528 468.655 1.847 0.642 3.088 AP-2-D 112.039 555.076 505.205 343.297 1.715 0.804 2.686 AP-1-D 128.485 464.956 514.582 419.836 1.483 0.665 2.317 DP-M 154.377 275.458 533.794 488.772 0.58 −0.618 1.78 AP-2-TE 105.94 137.353 537.165 508.375 0.114 −1.072 1.313 AP-1-M 182.987 163.186 474.7 525.987 −0.206 −1.381 0.998 AP-1-TE 241.495 190.13 501.361 502.648 −0.351 −1.557 0.895 AP-2-M 201.415 113.844 488.782 540.889 −0.555 −1.623 0.532 TH, thickness (μm). Ranking reflects decreasing effectiveness.Figure 9. EF for all protocol–photosensitizer combinations. Bars represent EF values computed from LME-predicted means, and error barsindicate 95% CIs obtained via bootstrap resampling. Positive EF values indicate a greater relative increase in shear modulus compared with thickness change, whereas negative values indicate an unfavorable trade-off. CIs crossing zero denote effects that are not statistically distin- guishable from a neutral balance between mechanical strengthening and structural change.approach accounts for uncertainty in both contributing variables and appropriately propagates it through the nonlinear EF formulation. Estimated marginal means were computed using equal weighting across angular positions, ensuring that each semimeridian contributed equally to the global corneal estimate. The EF values for all protocol–photosensitizer combinations are presented in Table 4 and illustrated in Figure 9. The DP–D combination exhibits the highest EF (EF = 4.21; 95% confidence interval [CI], 3.05–5.49), indicating a substantially greater relative increase in shear modulus compared with thickness change. This is followed by DP–TE (EF = 1.85; 95% CI, 0.64–3.09), AP-2–D (EF = 1.72; 95% CI, 0.80– 2.69), and AP-1–D (EF = 1.48; 95% CI, 0.67–2.32), all demonstrating favorable biomechanical strengthening relative to structural alteration. Moderate effectiveness is observed for DP–M (EF =0.58; 95% CI, −0.62 to 1.78) and AP-2–TE (EF = 0.11; 95% CI, −1.07 to 1.31), where CIs span zero, indicat- ### Page 16 Corneal Stiffening in Accelerated UV-A Crosslinking TVST | July 2026 | Vol. 15 | No. 7 | Article 14 | 16ing that the balance between stiffness gain and thick- ness change is not statistically distinguishable from a neutral effect. In contrast, AP-1–M (EF = −0.21; 95% CI, −1.38 to 1.00), AP-1–TE (EF = −0.35; 95% CI,−1.56 to 0.90), and AP-2–M (EF = −0.56; 95% CI,−1.62 to 0.53) yield negative point estimates; however, their wide CIs also overlap zero, suggesting that these apparently unfavorable trade-offs are not robust. Overall, these results indicate that the DP proto- col, particularly when combined with D or TE photo- sensitizers, achieves the most favorable biomechanical efficiency. Importantly, the inclusion of bootstrap- derived CIs demonstrates that only a subset of combi- nations, primarily DP–D and, to a lesser extent, DP–TE and AP-2–D, show consistently positive and well-defined effects, while other combinations exhibit substantial uncertainty. This finding highlights the importance of accounting for variability when inter- preting the EF, because point estimates alone may overstate the strength or direction of biomechanical effects.DiscussionIn this study, we used ACUS-OCE for nondestruc- tive, noncontact, and in situ biomechanical charac- terization of ex vivo rabbit corneas across distinct CXL phases. The ACUS-OCE enables rapid, multi- meridian measurements of corneal wave propagation and corneal thickness along 16 angles spanning 360° around the corneal apex within a 2-second scan acquisi- tion, allowing spatially resolved mapping at the Virgin, Post-RB (after 30 minutes of RB soaking), and Post- CXL (after 30, 9, or 3 minutes of UV-A irradiation and RB soaking) phases without interrupting the CXL procedure. Moreover, the ultrasonic excitation (500 kHz) provides consistent, repeatable stimulation of A0 Lamb waves (2.5 kHz) without inducing dehydration, tissue disruption, or macroscopic deformation, outper- forming air-pulse OCE,40–42 contact-based elastogra- phy,43–45 uniaxial/biaxial extensiometry,46,47 and clini- cal tonometry, such as Corvis and Oculus Response Analyser.48,49 In wave-based OCE50 in the cornea, symmetric or antisymmetric zero-order wave modes are typically measured, and phase speed and thickness are used to estimate the shear modulus shear modulus via the mRLFE.36,41,43 This model allows for a robust estima- tion of the elastic part of shear modulus (and not the viscoelasticity) at a specific frequency, even in the absence of full phase speed dispersion measurements. Provided that the acoustic pressure field is incident normal to the corneal apex, the antisymmetric zero- order mode is selectively excited—as in this study—as demonstrated by Duvvuri et al.41 Because the mRLFE formulation incorporates both phase velocity and local thickness as independent variables, it inherently accounts for geometric variations, ensuring a consis- tent biomechanical interpretation of the shear modulus across different corneal thicknesses. Given the preferen- tial lamellar organization of collagen fibers, the cornea is modeled as a transversely isotropic medium. In this configuration, the out-of-plane shear modulus— representing shear deformations in the x–z and y–zplanes—differs from the in-plane shear modulus (μ or the x–y plane), which is more closely associated with tensile stiffness.9,41 Because the Lamb waves gener- ated in this study primarily induce shear deforma- tions perpendicular to the corneal surface, the calcu- lated modulus corresponds to the out-of-plane shear modulus (G) as defined by Pitre et al.9 Our results demonstrate significant biomechanical effects of corneal CXL in rabbit corneas, and differ- ences across different RB photosensitizers (D, M, and TE) and protocols (DP, AP-1, and AP-2), assessed at distinct treatment phases (Virgin, Post-RB, and Post- CXL). In all cases, changes in shear modulus were primarily driven by photosensitizer (F = 40.63) and, to a lesser extent, by protocol alone (F = 13.98), and protocol × photosensitizer (F = 5.84), indicat- ing that the biomechanical response of the cornea is not solely dependent on individual factors, but is modulated by their specific combinations. Among all combinations, DP-D (Dresden protocol with dextran- based RB photosensitizer) produced the most signifi- cant corneal stiffening effect Post-CXL, with a mean increase in shear modulus of 1054.9 kPa (P < 0.001), but also induced the greatest corneal thinning, with a 186-μm decrease in thickness (P < 0.001). These findings are consistent with previously reported results using various elastography techniques.5,8,10,12 When comparing the impact of CXL photosen- sitizers only during the RB soaking phase (Virgin– Post-RB phase), we found that the D photosensitizer produced the most significant shear modulus increases and thickness decrease across all protocols (DP, AP- 1, and AP-2), with no significant changes for M or TE. This finding highlights the biomechanical impact of dextran-based RB photosensitizers in the cornea in addition to the corneal thinning effect due to its high osmolarity that draws fluid out of the corneal stroma. Therefore, the dehydration caused by the D formula- tion could produce an increase of the shear modulus of the extracellular matrix of the stroma, associated with an increase in the propagation phase speed of A0 Lamb waves in cornea.44 ### Page 17 Corneal Stiffening in Accelerated UV-A Crosslinking TVST | July 2026 | Vol. 15 | No. 7 | Article 14 | 17We found the APs (AP-1 and AP-2) to be less effective in stiffening the cornea, compared with the standard DP, rendering some RB photosensitizer formulations ineffective when applied in an accelerated regime. For example, the epi-off M photosensitizer did not produce a significant increase in shear modulus, whereas epi-on TE only achieved significant stiffening if applied in the DP protocol. Similarly, the D photo- sensitizer achieved significant stiffening only with the AP-1 and AP-2 protocols after RB soaking (Post-RB). Interestingly, a decrease in shear modulus was observed during the irradiation phase (post-CXL) for the AP-2- D group. This reduction can be attributed to corneal swelling, which likely exerted a more dominant biome- chanical effect than the CXL process itself during the 3-minute irradiation period. This is supported by the significant thickness increase reported for AP-2-D (see Figs. 8c, 8d) and aligns with previous findings by Singh et al.,51 who demonstrated that increased hydra- tion levels lead to a measurable reduction in the corneal shear modulus. In this study, the APs exhibited a trend toward increased shear modulus, although these changes did not reach statistical significance. This lower stiffening efficacy compared with the standard DP is consistent with ex vivo observations using Brillouin microscopy,16 inflation test,18 stress-strain extensome- try,52,53 and scanning acoustic microscopy54; and with clinical observations14,15,17 as reported in the litera- ture. While comparing in vivo human data with ex vivo OCE measurements requires caution due to differ- ences in hydration and IOP, both appear to reflect the same underlying biomechanical trends. The reduced efficiency of the APs may result from rapid oxygen consumption that outpaces stromal diffusion, leading to oxygen depletion and a shift from aerobic to a less efficient anaerobic CXL mechanism. In addition, shorter irradiation times limit RB diffusion and can accelerate RB photobleaching, thereby reducing the effective CXL depth.21 Furthermore, as Lamb wave speed represents a depth-averaged measurement of corneal stiffness, it may inherently underestimate the peak stiffness of the most superficial layers, which are primarily affected by the limited penetration of APs. To address this, stratified corneal models—similar to those used by Bekesi et al.55 to differentiate the penetra- tion depths of Rose Bengal vs. RB CXL—could be applied to OCE data in the future. However, for the clinical purpose of preventing ectasia progression, the effective bulk stiffening provided by depth-averaged measurements may serve as a more robust predictor of overall structural stability than localized anterior stiff- ness alone. When comparing the biomechanical effects of epi- on (TE) vs. epi-off (D, M) photosensitizers the DP- TE produced significant increase in shear modulus (G = 344.6 kPa; P < 0.01), demonstrating the potential of BKC-based transepithelial formulations to stiffen the cornea without inducing significant thinning or epithelial disruption. Similar behavior was previously reported by Armstrong et al.,24 who observed greater—but not statistically significant— stiffening using a BKC-EDTA epi-on RB formula- tion compared with the standard DP and other epi- on protocols, as measured using OCE. In contrast, no significant increase in shear modulus was detected for APs (AP-1, and AP-2) when combined with TE photo- sensitizer. Because the efficiency of the CXL proce- dure depends in part on the diffusion and penetration of RB, and this varies with protocols and photosensi- tizer formulation (epi-on and epi-off), depth-resolved methods of shear modulus should provide further mechanistic insights on CXL. Although in this study we report average shear modulus in each corneal merid- ians, possible depth-resolved reconstruction of shear modulus in cornea after CXL is being explored by Regnault et al.56 using wave-based OCE. Corneal thinning was observed following full CXL (Virgin–Post-CXL) across all protocols (DP, AP-1, and AP-2) when using the D photosensitizer, consis- tent with known effects associated with 20% dextran content in the RB solution. In contrast, corneal thick- ening was observed in the AP-1 and AP-2 proto- cols when using the M photosensitizer. This response is likely due to endothelial cell disruption, which facilitates fluid influx and stromal swelling, as previ- ously reported by other groups,51,57 although the extent of the differences of this effect in corneas in in vivo or ex vivo merits further investigation. Such swelling may contribute to corneal softening and could partially explain the reduced stiffening efficacy observed with the M photosensitizer in our measurements. Meanwhile, the TE photosensitizer did not induce significant changes in corneal thickness across any of the three protocols—DP, AP-1, and AP-2—underscoring the advantage of transepithelial (epi-on) treatments in maintaining structural corneal integrity. Importantly, it should be noted that corneal thinning induced by dextran-containing solutions is a transient effect during treatment that is typically not observed in clinical follow-up periods several weeks postoperatively.22 Within the context of our study, the proposed EF incorporates these transient thickness changes to provide a standardized metric for compar- ing biomechanical outcomes across different experi- mental groups. However, we acknowledge that the EF serves primarily as a comparative tool for the final state and does not fully account for how stromal hydra- tion or compactness at the time of irradiation might ### Page 18 Corneal Stiffening in Accelerated UV-A Crosslinking TVST | July 2026 | Vol. 15 | No. 7 | Article 14 | 18fundamentally alter the photochemical efficiency of the CXL process itself. Although these ex vivo findings offer valuable insights into protocol-dependent stiffen- ing, direct protocol optimization for clinical use would require further validation to account for the dynamic biological environment and oxygen-limited reactions present in vivo. The baseline results of this study reveal that the rabbit cornea exhibits significant angular dependence of the shear modulus in the Virgin phase. This hetero- geneity likely reflects the complex underlying lamellar organization and preferential collagen fiber orientation characteristic of the corneal stroma. However, most CXL protocol–photosensitizer combinations resulted in a uniform stiffening effect across all 16 measured semimeridians (P > 0.05), with exceptions observed when using D photosensitizer. Specifically, DP-D and AP-2-D led to nonuniform changes in shear modulus during both the RB soaking (P < 0.01) and UV irradi- ation (P < 0.05) phases. The observation that CXL may act as a global reinforcing agent to normalize inherent asymmetries remains preliminary. The specific mecha- nisms by which different photosensitizing agents inter- act with local stromal variations to either maintain or disrupt mechanical symmetry require further explo- ration in greater detail to fully understand the spatial impact of the treatment. Some limitations in this study warrant consider- ation. Firstly, while ex vivo models may not fully replicate in vivo hydration dynamics, previous studies comparing CXL outcomes in both settings have shown that the general stiffening trends—particularly for the DP—remain consistent across both environments as shown in Bekesi et al.58 Consequently, despite the inherent differences in physiological conditions, our ex vivo findings provide a reliable perspective on the relative efficacy of various protocol-photosensitizer combinations. A second limitation is related to corneal dehydration. While balanced salt solution was applied via pipette to maintain corneal hydration between measurements, we anticipate minimal biomechanical softening compared with Phosphate-Buffered Saline (PBS) -based protocols.51 Unlike PBS, the divalent cations in balanced salt solution support the struc- tural integrity of cellular tight junctions and metabolic pumps,59 which helps stabilize the stroma against the rapid swelling or breakdown often seen in ex vivo models. Future studies may benefit from humidity- controlled chambers to further standardize the hydra- tion state across all experimental phases.60 A third limitation involves the inherent physics of Lamb wave propagation in the cornea. Lamb wave-based OCE methods provide an effective shear modulus shear modulus for the entire corneal thickness rather than a depth-resolved map, which may complicate the assess- ment of protocols that induce highly superficial stiff- ening (e.g., accelerated CXL). However, analytical modeling of bilayer media by Regnault et al.,56 consist- ing of a stiff layer over a soft layer demonstrates that Lamb wave speed is determined by the volume- weighted contribution of both layers. Specifically, a 3- fold increase in shear modulus constrained only to the anterior 10% of the cornea would still produce a statis- tically significant decrease in the measured wavenum- ber (k) and a corresponding increase in the phase velocity compared with untreated controls. Thus, while our current mRLFE model approximates the cornea as a homogeneous medium, it remains sensitive to the total structural reinforcement provided by thin crosslinked layers. Future work involving multilayered analytical models, as explored by Regnault et al.,56 will focus on reconstructing depth-dependent shear moduli to further refine the assessment of accelerated CXL protocols. Other limitations of the study lie in structural assumptions of the cornea. Although all eyes were systematically aligned with the anatomical superior– inferior meridian, asymmetric patterns of shear modulus may arise from the nonuniform stress distributions inherent to corneal anatomy or from experimental artifacts introduced by the needles in the IOP control system. Although we minimized the needle gauge to reduce such effects, our analysis of shear modulus and corneal thickness per meridian was specifically designed to limit the impact of both physi- ological and artefactual asymmetries. Additionally, we noted inconsistencies in the Virgin to Post-RB corneal thickness variations within the same photosensitizer groups. For the dextran-based cases, although thinning was universal, the magnitude varied between batches in each protocol. The observed variability in corneal thinning after RB-dextran instillation is consistent with the wide range of osmotic effects reported in previous ex vivo literature. For instance, whereas De Paula et al.61 reported a 12.9% decrease in thickness, prior studies from our group using Scheimpflug62 and OCT63 imaging have observed thinning as high as 47% to 48%. This fluctuation likely stems from varying degrees of baseline edema in slaughterhouse-acquired samples, where more hydrated corneas may exhibit a nonlinear osmotic response to dextran. Similarly, although the M group generally exhibited the expected thickening associated with HPMC-based solutions (previously reported as a 9.1% increase),61 one case (DP-M) unexpectedly showed a decrease. Finally, the reported shear modulus values align with the expected order of magnitude for rabbit ocular tissue, which exhibits a fundamental discrepancy between in-plane ### Page 19 Corneal Stiffening in Accelerated UV-A Crosslinking TVST | July 2026 | Vol. 15 | No. 7 | Article 14 | 19tensile stiffness (measured in megapascals) and out-of- plane shear stiffness (measured in kilopascals) due to its transversely isotropic nature. As demonstrated by Villegas et al.26 using identical instrumentation, our results are consistent with previously reported rabbit data, and the higher magnitudes found in human studies such as that by Kirby et al.8 are attributable to species-specific differences in collagen density and tissue preparation. In summary, this study provided in-depth compar- ative information of the biomechanical impact of UV-A RB-based CXL in ex vivo and in situ rabbit corneas (pressurized at 15 mm Hg) when using the DP and APs (AP-1 and AP-2), for different RB epi- off (D, M) and epi-on (TE) photosensitizers at three treatment phases (Virgin, Post-RB, and Post-CXL). We explored stiffening (increase of shear modulus) and corneal thinning effects at each phase for each protocol–photosensitizer combination. The proposed EF in Equation 2 and the EF results of Figure 9 shows that DP-TE offers a good compromise of both effects (EF = 1.847), while preserving corneal integrity and improving post-treatment recovery. Although the proposed EF metric quantifies the balance between corneal geometric changes and stiffening effects, it does not directly take into account corneal parameters such as hydration and compactness. Further analysis on how this EF is preserved over time and its capabilities of halting the progression of keratoconus is needed with in vivo clinical studies. The DP-D continues to have the strongest impact on the stiffening of corneas by the cost of corneal thinning, which makes it the most effec- tive treatment (EF = 4.208). Unfortunately, accord- ing to our study findings, APs do not provide signif- icant stiffening of the cornea, with the exception of AP-1-D, in which the stiffening is mostly produced in the RB-soaking phase and not during UV-A irradiation. Future preclinical research can explore the effect of IOP on the shear modulus values, as well as the real- time biomechanical response of the cornea to CXL, by acquiring time-resolved OCE measurements through- out the course of different protocols and RB treat- ments to generate time-dependent stiffness profiles and investigate potential nonlinear effects related to UV- A irradiation dose and duration. Most important, the speed, safety, and adaptability of the ACUS-OCE technique enables the in vivo monitoring of corneal biomechanics both in animal models and in patients. In this context, future clinical studies can use ACUS- OCE to track the evolution of the EF throughout the postoperative recovery period and provide insight into the long-term biomechanical efficacy of various CXL strategies10,30,31 in keratoconus patients along many years, as shown in the study by Felter et al.64 The ability of ACUS-OCE to spatially resolve the corneal shear modulus in patients opens the possibility of selectively irradiating regions with differing stiffness, thereby enabling patient-specific customization of the CXL procedure and minimizing stiffness heterogeneity across the cornea.ConclusionsThis study shows important biomechanical differ- ences between combinations of RB photosensitizers (D, M, and TE) and protocols (DP, AP-1, and AP- 2) during CXL of ex vivo rabbit corneas. We found that APs did not produce significant corneal stiffen- ing for any RB photosensitizer, except by AP-1-D. Although the DP-D produces the strongest corneal stiffening effect, M treatment produces the lowest stiffening in all protocols attributed to absence of dextran. Interestingly, DP-D and DP-TE produce the greatest balance between shear modulus increase and corneal thickness decrease during ex vivo condi- tions, suggesting a possible research avenue for in vivo studies using DP-TE as a potential treat- ment with great compromise between biomechani- cal impact, corneal integrity, and faster postoperative recovery.AcknowledgmentsSupported by the Spanish Government FIS2017- 84753-R, PID2020-115191RB-I00, IND2019/BMD- 17262; Horizon 2020 European Project Imcustomeye (H2020-ICT-2017 Ref. 779960); NIH NEI P30EY 001319; Unrestricted Funds Research to Prevent Blindness; Ayudas Ramón y Cajal (RYC2022-037910- I); Ayudas Atracción de Talento RyC (20245AT005); Fundacja na rzecz Nauki Polskiej (MAB/2019/12); National Agency for Academic Exchange, Polish Government (NAWA ULAM/2020/1/00176); and Peruvian CONCYTEC-PROCIENCIA N° PE50108 2168-2023; PE501093888-2024 and N° PE501098889- 2025; Wellcome Trust Early Career Award 300307/Z/23/Z; OPTICA Foundation Challenge Award (2023). Disclosure: F. Zvietcovich, (P); A. Varea-Bejar, None; V. Pleshcheva, None; A. Curatolo, None;J.S. Birkenfeld, None; S. Marcos, (P) ### Page 20 Corneal Stiffening in Accelerated UV-A Crosslinking TVST | July 2026 | Vol. 15 | No. 7 | Article 14 | 20References1. Vazirani J, Basu S. Keratoconus: current per- spectives. Clin Ophthalmol (Auckland, NZ). 2013;7:2019–2030. 2. Roberts CJ, Dupps WJJ. Biomechanics of corneal ectasia and biomechanical treatments. J Cataract Refract Surg. 2014;40:991–998. 3. Wollensak G, Spoerl E, Seiler T. Riboflavin/ ultraviolet-A–induced collagen crosslinking for the treatment of keratoconus. Am J Ophthalmol. 2003;135:620–627. 4. Cherfan D, Verter EE, Melki S, et al. Collagen cross-linking using rose Bengal and green light to increase corneal stiffness. Invest Ophthalmol Vis Sci. 2013;54:3426–3433. 5. Blackburn BJ, Rollins AM, Dupps WJ, Jr. Biome- chanics of ophthalmic crosslinking. Transl Vis Sci Technol. 2021;10:8. 6. Tang Y, Jiang J, Yang C, Yin H, Fei Y. Com- parison of the 1-year efficacy of epithelium- off and transepithelial accelerated corneal colla- gen crosslinking in adolescent keratoconus. Eye. 2025;39:1967–1974. 7. Juthani VV, Chuck RS. Corneal crosslinking in refractive corrections. Transl Vis Sci Technol. 2021;10:4. 8. Kirby MA, Pelivanov I, Regnault G, et al. Acous- tic micro-tapping optical coherence elastography to quantify corneal collagen cross-linking: an ex vivo human study. Ophthalmol Sci. 2023;3:100257. 9. Pitre JJ, Jr, Kirby MA, Li DS, et al. Nearly- incompressible transverse isotropy (NITI) of cornea elasticity: model and experiments with acoustic micro-tapping OCE. Sci Rep. 2020;10: 12983. 10. Zvietcovich F, Nair A, Singh M, Aglyamov SR, Twa MD, Larin KV. In vivo assessment of corneal biomechanics under a localized cross-linking treat- ment using confocal air-coupled optical coherence elastography. Biomed Opt Express. 2022;13:2644– 2654. 11. Blackburn BJ, Gu S, Ford MR, et al. Noninva- sive assessment of corneal crosslinking with phase- decorrelation optical coherence tomography. Invest Ophthalmol Vis Sci. 2019;60:41–51. 12. Zhao Y, Zhu Y, Yan Y, et al. In vivo evaluation of corneal biomechanics following cross-linking surg- eries using optical coherence elastography in a rab- bit model of keratoconus. Transl Vis Sci Technol. 2024;13:15. 13. Ge G, Tavakol B, Usher D, Adler D, Rolland J, Parker K. Assessing corneal cross-linking with reverberant 3D optical coherence elastography.J Biomed Opt. 2022;27:026003. 14. Shajari M, Kolb CM, Agha B, et al. Com- parison of standard and accelerated corneal cross-linking for the treatment of keratoconus: a meta-analysis. Acta Ophthalmol. 2019;97:e22– e35. 15. Sadoughi MM, Einollahi B, Baradaran-Rafii A, Roshandel D, Hasani H, Nazeri M. Acceler- ated versus conventional corneal collagen cross- linking in patients with keratoconus: an intrapa- tient comparative study. Int Ophthalmol. 2018;38: 67–74. 16. Webb JN, Su JP, Scarcelli G. Mechanical out- come of accelerated corneal crosslinking evaluated by Brillouin microscopy. J Cataract Refract Surg. 2017;43:1458–1463. 17. Sherif AM. Accelerated versus conventional corneal collagen cross-linking in the treatment of mild keratoconus: a comparative study. Clin Ophthalmol. 2014;8:1435–1440. 18. Bao F, Zheng Y, Liu C, et al. Changes in corneal biomechanical properties with different corneal cross-linking irradiances. J Refract Surg. 2018;34:51–58. 19. Bunsen R, Roscoe H. Photochemische Unter- suchungen. Annalen der Physik. 1857;176:43–88. 20. Seiler TG, Komninou MA, Nambiar MH, Schuerch K, Frueh BE, Büchler P. Oxygen kinetics during corneal cross-linking with and without supplementary oxygen. Am J Ophthalmol. 2021;223:368–376. 21. Raiskup F, Herber R, Lenk J, Pillunat LE, Spoerl E. Crosslinking with UV-A and riboflavin in pro- gressive keratoconus: From laboratory to clinical practice – developments over 25 years. Prog Retin Eye Resv 2024;102:101276. 22. Bradford S, Luo S, Brown D, Juhasz T, Jester J. A review of the epithelial and stromal effects of corneal collagen crosslinking. Ocul Surf. 2023;30:150–159. 23. Nath S, Shen C, Koziarz A, et al. Transepithelial versus epithelium-off corneal collagen cross- linking for corneal ectasia: a systematic review and meta-analysis. Ophthalmology. 2021;128:1150– 1160. 24. Armstrong BK, Lin MP, Ford MR, et al. Bio- logical and biomechanical responses to tradi- tional epithelium-off and transepithelial riboflavin- UVA CXL techniques in rabbits. J Refract Surg. 2013;29:332–341. 25. Touboul D, Gennisson JL, Nguyen TM, et al. Supersonic shear wave elastography for the in vivo evaluation of transepithelial corneal col- ### Page 21 Corneal Stiffening in Accelerated UV-A Crosslinking TVST | July 2026 | Vol. 15 | No. 7 | Article 14 | 21lagen cross-linking. Invest Ophthalmol Vis Sci. 2014;55:1976–1984. 26. Villegas L, Zvietcovich F, Marcos S, Birken- feld JS. Revealing regional variations in scleral shear modulus in a rabbit eye model using multi- directional ultrasound optical coherence elastogra- phy. Sci Rep. 2024;14:21010. 27. Zvietcovich F, Varea A, Alejandre-Alba N, Mer- ayo J, Birkenfeld J, Marcos S. Optical coherence elastography for customized eye care: a success case of clinical translation (Conference Presenta- tion). Proc SPIE PC12381, Optical Elastography and Tissue Biomechanics X, PC123810R (15 Mar 2023). https://doi.org/10.1117/12.2654746. 28. Marcos S, Varea A, Birkenfeld J, Zvietcovich F. Biomechanical assessment of corneal colla- gen crosslinking treatments using ultrasonic wave- based optical coherence elastography. Invest Oph- thalmol Vis Sci. 2023;64:1693–1693. 29. Zvietcovich F, Nair A, Ambekar YS, et al. Con- focal air-coupled ultrasonic optical coherence elas- tography probe for quantitative biomechanics. Opt Lett. 2020;45:6567–6570. 30. Zvietcovich F, Varea A, Alejandre-Alba N, Merayo-Lloves J, Birkenfeld J, Marcos S. Detect- ing subclinical keratoconus by mapping corneal biomechanics using wave-based optical coherence elastography. Proc. SPIE PC12360, Ophthalmic Technologies XXXIII, PC123600O (17 Mar 2023). https://doi.org/10.1117/12.2649414. 31. Zvietcovich F, Birkenfeld J, Varea A, Gonza- lez AM, Curatolo A, Marcos S. Multi-meridian wave-based corneal optical coherence elastogra- phy in normal and keratoconic patients. Invest Ophthalmol Vis Sci. 2022;63:2380–A0183–2380– A0183. 32. Curatolo A, Birkenfeld JS, Martinez-Enriquez E, et al. Multi-meridian corneal imaging of air-puff induced deformation for improved detection of biomechanical abnormalities. Biomed Opt Express. 2020;11:6337–6355. 33. Villegas L, Zvietcovich F, Varea A, Birkenfeld JS, Marcos S. Spatial characterization of scleral biomechanics in ex vivo rabbit eyes using multi- meridian optical coherence elastography. Invest Ophthalmol Vis Sci. 2022;63:1454–F0412–1454– F0412. 34. Bronte-Ciriza D, Birkenfeld JS, de la Hoz A, et al. Estimation of scleral mechanical properties from air-puff optical coherence tomography. Biomed Opt Express. 2021;12:6341–6359. 35. Zvietcovich F, Larin KV. Wave-based optical coherence elastography: the 10-year perspective.Prog Biomed Eng. 2021;4:012007. 36. Han Z, Li J, Singh M, et al. Optical coherence elas- tography assessment of corneal viscoelasticity with a modified Rayleigh-Lamb wave model. J Mech Behav Biomed Mater. 2017;66:87–94. 37. Han Z, Aglyamov SR, Li J, et al. Quantita- tive assessment of corneal viscoelasticity using optical coherence elastography and a modified Rayleigh–Lamb equation. J Biomed Opt. 2015;20: 020501. 38. Zvietcovich F, Pongchalee P, Meemon P, Rol- land JP, Parker KJ. Reverberant 3D optical coher- ence elastography maps the elasticity of individual corneal layers. Nat Commun. 2019;10:4895. 39. Oberg AL, Mahoney DW. Linear mixed effects models. In: Ambrosius WT (ed), Topics in Bio- statistics. Totowa, NJ: Humana Press; 2007:213– 234. 40. Wang S, Larin KV. Noncontact depth-resolved micro-scale optical coherence elastography of the cornea. Biomed Opt Express. 2014;5:3807– 3821. 41. Duvvuri C, Singh M, Lan G, Aglyamov SR, Larin KV, Twa MD. Determinants of human corneal mechanical wave dispersion for in vivo optical coherence elastography. Transl Vis Sci Technol. 2025;14:26. 42. Lan G, Aglyamov SR, Larin KV, Twa MD. In vivo human corneal shear-wave optical coherence elas- tography. Optom Vis Sci. 2021;98:58–63. 43. Zvietcovich F, Nair A, Singh M, Aglyamov SR, Twa MD, Larin KV. Dynamic optical coherence elastography of the anterior eye: understanding the biomechanics of the limbus. Invest Ophthalmol Vis Sci. 2020;61:7. 44. Ramier A, Eltony AM, Chen Y, et al. In vivo measurement of shear modulus of the human cornea using optical coherence elastography. Sci Rep. 2020;10:17366. 45. Ramier A, Tavakol B, Yun S-H. Measuring mechanical wave speed, dispersion, and viscoelas- tic modulus of the cornea using optical coherence elastography. Opt Express. 2019;27:16635–16649. 46. Søndergaard AP, Ivarsen A, Hjortdal J. Corneal resistance to shear force after UVA-riboflavin cross-linking. Invest Ophthalmol Vis Sci. 2013; 54:5059–5069. 47. Chang SH, Mohammadvali A, Chen KJ, et al. The relationship between mechanical properties, ultra- structural changes, and intrafibrillar bond forma- tion in corneal UVA/riboflavin cross-linking treat- ment for keratoconus. J Refract Surg. 2018;34:264– 272. 48. Jabbarvand M, Moravvej Z, Shahraki K, et al. Corneal biomechanical outcome of collagen cross- ### Page 22 Corneal Stiffening in Accelerated UV-A Crosslinking TVST | July 2026 | Vol. 15 | No. 7 | Article 14 | 22linking in keratoconic patients evaluated by Corvis ST. Eur J Ophthalmol. 2021;31:1577–1583. 49. Viswanathan D, Kumar NL, Males JJ, Graham SL. Relationship of structural characteristics to biomechanical profile in normal, keratoconic, and crosslinked eyes. Cornea. 2015;34:791–796. 50. Kirby MA, Pelivanov I, Song S, et al. Opti- cal coherence elastography in ophthalmology.J Biomed Opt. 2017;22:1–28. 51. Singh M, Han Z, Li J, et al. Quantifying the effects of hydration on corneal stiffness with noncontact optical coherence elastography. J Cataract Refract Surg. 2018;44:1023–1031. 52. Herber R, Francis M, Spoerl E, Pillunat LE, Raiskup F, Sinha Roy A. Comparison of waveform-derived corneal stiffness and stress- strain extensometry-derived corneal stiffness using different cross-linking irradiances: an experimen- tal study with air-puff applanation of ex vivo porcine eyes. Graefes Arch Clin Exp Ophthalmol. 2020;258:2173–2184. 53. Hammer A, Richoz O, Arba Mosquera S, Tabibian D, Hoogewoud F, Hafezi F. Corneal biomechan- ical properties at different corneal cross-linking (CXL) irradiances. Invest Ophthalmol Vis Sci. 2014;55:2881–2884. 54. Beshtawi IM, Akhtar R, Hillarby MC, et al. Biomechanical properties of human corneas following low- and high-intensity collagen cross-linking determined with scanning acous- tic microscopy. Invest Ophthalmol Vis Sci. 2013;54:5273–5280. 55. Bekesi N, Kochevar IE, Marcos S. Corneal biome- chanical response following collagen cross-linking with Rose Bengal–green light and riboflavin-UVA.Invest Ophthalmol Vis Sci. 2016;57:992–1001. 56. Regnault G, Kirby MA, Wang RK, Shen TT, O’Donnell M, Pelivanov I. Possible depth-resolved reconstruction of shear moduli in the cornea fol- lowing collagen crosslinking (CXL) with optical coherence tomography and elastography. Biomed Opt Express. 2023;14:5005–5021. 57. Vantipalli S, Li J, Singh M, Aglyamov SR, Larin KV, Twa MD. Effects of thickness on corneal biomechanical properties using optical coherence elastography. Optom Vis Sci. 2018;95:299–308. 58. Bekesi N, Gallego-Muñoz P, Ibarés-Frías L, et al. Biomechanical changes after in vivo col- lagen cross-linking with Rose Bengal–green light and riboflavin-UVA. Invest Ophthalmol Vis Sci. 2017;58:1612–1620. 59. Doughty MJ. Evaluation of the effects of saline versus bicarbonate-containing mixed salts solu- tions on rabbit corneal epithelium in vitro. Oph- thalmic Physiol Opt. 1995;15:585–599. 60. Kling S, Marcos S. Effect of hydration state and storage media on corneal biomechanical response from in vitro inflation tests. J Refract Surg. 2013;29:490–497. 61. De Paula TAA, Cresta FB, Alves MR. Compar- ative two-photon fluorescence microscopy analy- sis of riboflavin penetration in two different solu- tions: dextran and hydroxypropyl methylcellulose.Clin Ophthalmol. 2020;14:1867–1874. 62. Kling S, Remon L, Pérez-Escudero A, Merayo- Lloves J, Marcos S. Corneal biomechanical changes after collagen cross-linking from porcine eye inflation experiments. Invest Ophthalmol Vis Sci. 2010;51:3961–3968. 63. Dorronsoro C, Pascual D, Pérez-Merino P, Kling S, Marcos S. Dynamic OCT measurement of corneal deformation by an air puff in normal and cross-linked corneas. Biomed Opt Express. 2012;3:473–487. 64. Felter E, Khoramnia R, Friedrich M, Son HS, Auf- farth GU, Augustin VA. Biomechanical changes following corneal crosslinking in keratoconus patients. Graefes Arch Clin Exp Ophthalmol. 2024;262:3635–3642.