Introduction: The Biomechanical Challenge in Refractive Surgery

Refractive surgeries such as LASIK, PRK, and SMILE have transformed the correction of myopia, hyperopia, and astigmatism. While these procedures are widely successful, their outcomes depend critically on how the eye’s tissues, especially the cornea, respond mechanically to surgical cuts and tissue removal. A poor mechanical response can lead to complications like corneal ectasia, irregular astigmatism, or delayed wound healing. Understanding and predicting these responses through computational modeling has become an essential pillar for improving surgical safety and precision.

The eye is a pressurized, viscoelastic organ. The cornea—its primary refractive element—must maintain a precise curvature and thickness while bearing intraocular pressure (IOP) and withstanding surgical trauma. Modeling the mechanical response allows surgeons to simulate outcomes, evaluate risks, and customize procedures for individual patients. This article explores the principles, methods, and clinical relevance of modeling the eye’s mechanical behavior during and after refractive surgery.

Corneal Structure and Mechanical Properties

Layered Architecture

The human cornea consists of five layers: epithelium, Bowman’s layer, stroma, Descemet’s membrane, and endothelium. The stroma constitutes about 90% of corneal thickness and is composed of collagen fibrils (predominantly types I and V) arranged in orthogonal lamellae. This structure provides both strength and transparency. The lamellar orientation varies regionally—more aligned in the central cornea and more interwoven near the limbus—contributing to anisotropic mechanical behavior.

Viscoelastic and Nonlinear Behavior

Corneal tissue exhibits time-dependent (viscoelastic) and nonlinear (hyperelastic) responses. Under low strain, the collagen crimp straightens; at higher strains, the fibrils themselves stretch. This nonlinearity is captured by hyperelastic material models such as the Holzapfel–Gasser–Ogden (HGO) or Mooney–Rivlin formulations. Viscoelasticity is important for modeling dynamic events like suction ring application or laser-induced tissue deformation. Additionally, corneal hydration significantly affects stiffness—edematous corneas are softer, while dehydration increases modulus.

Anisotropy and Fiber Orientation

Collagen fiber orientation introduces anisotropy. Models must account for preferential directions—especially the nasal-temporal and superior-inferior quadrants. Small-angle light scattering (SALS) and second harmonic generation (SHG) microscopy have quantified these orientations, enabling patient-specific fiber maps.

Refractive Surgery Procedures and Mechanical Impact

LASIK

In LASIK, a femtosecond laser creates a corneal flap (typically 100–110 µm thick), which is lifted after laser ablation of the underlying stroma. The flap weakens the anterior stroma, redistributing stress. The hinge location and flap thickness modify postoperative corneal biomechanics. Thinner flaps preserve more structural integrity but may be harder to reposition.

PRK

PRK removes the epithelium and Bowman’s layer by mechanical scraping or laser, followed by excimer laser ablation on the anterior stroma. No flap is created, which eliminates flap-related complications but exposes the tissue to longer healing times and earlier biomechanical weakening. The absence of a flap also means the entire anterior corneal stroma remains available for ablation, but the loss of Bowman’s membrane shifts stress more posteriorly.

SMILE

SMILE uses a femtosecond laser to create a lenticule within the stroma, which is then extracted through a small incision. The procedure spares the anterior stroma and Bowman’s layer, resulting in less alteration of corneal curvature and potentially greater biomechanical stability compared to LASIK. Studies suggest SMILE preserves more of the corneal tensile strength because the anterior lamellae remain mostly intact.

Modeling Approaches: From Analytical to Computational

Analytical Models

Early models used simple membrane or shell theories to estimate stress and strain. The Laplace law and thin-shell approximations provide quick predictions of central corneal curvature change after ablation. However, these models assume uniform thickness and isotropic, linear-elastic behavior—limitations that prevent accurate simulation of complex geometries or nonuniform ablations.

Finite Element (FE) Models

FE models dominate modern biomechanical simulations. A typical FE model of the eye includes the cornea, limbus, sclera, and sometimes the lens and ciliary body. The cornea is meshed with hexahedral or tetrahedral elements. Material models incorporate hyperelasticity (e.g., Ogden, HGO), viscoelasticity, and fiber orientation. Boundary conditions include prescribed IOP and constraint at the optic nerve. Ablation is simulated by removing elements or reducing thickness within a defined optical zone.

Key steps in FE modeling:

  • Geometry acquisition: From Scheimpflug imaging (e.g., Pentacam) or OCT to obtain anterior and posterior corneal surfaces, pachymetry, and corneal diameter.
  • Material parameter fitting: Using uniaxial or inflation tests on donor corneas. Recent studies use inverse FE methods to extract properties from clinical data like the Corvis ST or Ocular Response Analyzer.
  • Surgery simulation: Flap creation (partitioning the mesh), ablation (removing elements), and subsequent IOP loading.
  • Postoperative assessment: Prediction of curvature change, stress redistribution, and displacement maps.

Fluid-Structure Interaction (FSI) Models

FSI integrates fluid dynamics of the aqueous humor with solid mechanics of the cornea and sclera. These models capture the interaction between IOP fluctuations and tissue deformation. However, their computational cost is high, and they are primarily used in research settings to simulate dynamic events like eye rubbing or impact.

Machine Learning in Biomechanical Modeling

ML models (neural networks, Gaussian processes) can approximate patient-specific mechanical responses without solving full FE equations. They are trained on large datasets of FE simulations or clinical outcomes. While less physically interpretable, ML surrogates offer rapid predictions for real-time surgical planning. Hybrid approaches combine FE with ML to accelerate parameter estimation.

Factors Affecting the Mechanical Response

Intraocular Pressure and Corneal Thickness

IOP is the primary load on the cornea. A higher IOP increases stress in the remaining stroma after ablation, raising ectasia risk. Central corneal thickness (CCT) is a critical factor: thinner corneas (<500 µm) have less residual stroma, amplifying mechanical weakness. Models must account for individually measured IOP and CCT.

Residual Stromal Bed Thickness (RSBT)

RSBT is the thickness of untreated stroma beneath the ablation. Most guidelines recommend a minimum RSBT of 250–300 µm to maintain structural integrity. FE simulations confirm that RSBT below this threshold leads to dramatically higher stress concentrations at the ablation edge, especially in steep corneas.

Tissue Healing and Remodeling

Postoperative wound healing alters mechanical properties. Keratocyte activation, collagen deposition, and epithelial hyperplasia stiffen the corneal surface. Models that incorporate time-dependent healing (e.g., using an evolving stiffness field) better predict late outcomes like regression. The myofibroblast response can also change the local modulus by up to 50%.

Patient Age and Pre-Existing Conditions

Age reduces corneal stiffness due to collagen cross-linking changes. Older patients show less refractive stability, possibly related to altered biomechanical response. Conditions like keratoconus or forme fruste keratoconus exhibit significantly softer corneas; preoperative screening using biomechanical indices (e.g., Corvis’s deformation amplitude) is essential.

Clinical Applications of Biomechanical Modeling

Ectasia Risk Assessment

Corneal ectasia is one of the most feared complications. FE models can simulate the stress distribution in a candidate cornea after a planned ablation. If the stress exceeds a threshold (e.g., 0.2 MPa near the midpoint of the remaining stroma), the procedure is deemed high risk. Studies have validated these models against historical ectasia cases, showing >90% sensitivity.

Optimizing Flap and Ablation Parameters

Models help optimize flap thickness, hinge position, and ablation profile. For example, a thicker flap (120 µm vs 100 µm) redistributes more stress to the hinge area, which may reduce central stress but increase peripheral glare. Similarly, aspheric ablation profiles (e.g., Q-adjusted) produce flatter postoperative curvature, reducing stress risers compared to spherical profiles.

Customized Treatment for Astigmatism and Higher-Order Aberrations

Topography-guided and wavefront-optimized ablations aim to correct both spherocylindrical error and irregularities. FE models predict how local ablation depth variations affect curvature and stress uniformity. This is particularly important in eyes with asymmetric astigmatism or pellucid marginal degeneration, where standard nomograms may fail.

Postoperative IOP Measurement Adjustments

Corneal thinning after refractive surgery leads to underestimation of IOP by applanation tonometry. Biomechanical models can derive correction factors based on the change in corneal stiffness and thickness. Such adjustments are crucial for glaucoma screening in post-LASIK patients.

Challenges and Limitations of Current Models

Material Characterization Uncertainty

Corneal properties vary widely among individuals and even between regions of the same eye. Most models rely on population-averaged material constants, which may not capture individual response. Inverse characterization from noninvasive measurements (e.g., corneal hysteresis) is improving but still faces noise and resolution limits.

Computational Cost and Complexity

Detailed FE models with thousands of elements and nonlinear solvers require significant time (hours to days) to simulate one surgical scenario. This limits their use in routine clinical workflow. Parallel computing and reduced-order modeling are active areas of research.

Validation Against Clinical Outcomes

While FE predictions correlate with gross changes in central curvature, they often fail to predict higher-order aberrations or asymmetric healing. Prospective validation studies are sparse, partly because it is difficult to measure postoperative stress noninvasively. Emerging technologies like Brillouin microscopy may provide direct stiffness maps for validation.

Integration with Imaging and Surgical Systems

Translating models into operating room tools requires seamless data flow from diagnostic devices (e.g., OCT, topography) to simulation software. Standardized data formats and real-time solvers are needed but not yet widely adopted.

Recent Advances and Future Directions

Patient-Specific Modeling from Multimodal Imaging

Combining OCT-derived pachymetry maps with Schiempflug-derived anterior surface allows high-fidelity geometry. Some groups incorporate 3D collagen orientation from polarization-sensitive OCT or SHG microscopy, creating anisotropic material models unique to each patient. Early results show improved prediction of postoperative curvature compared to isotropic models.

Inverse Finite Element Methods

By fitting FE simulations to clinical biomechanical measurements (e.g., Corvis ST deformation or ORA waveform), researchers can extract patient-specific stiffness and viscoelastic parameters. This approach bypasses the need for ex vivo tests and enables adaptation to individual healing responses.

Machine Learning as a Surrogate for FE

Deep learning networks trained on thousands of FE simulations can predict outcomes in milliseconds. For instance, a convolutional neural network can take a pachymetry map and planned ablation profile as input and output the postoperative curvature and stress distribution. Such surrogates could be integrated into laser workstations for real-time risk assessment.

Coupling Biomechanics with Optics

Recent models link FE-predicted shape changes to ray-tracing simulations of the eye’s optical performance. This allows concurrent optimization of biomechanical stability and visual quality. For example, a simulated ablation can be tuned to minimize both stress concentration and spherical aberration simultaneously.

Role in Crosslinking and Combined Treatments

Corneal crosslinking (CXL) stiffens the cornea and is sometimes combined with refractive surgery (e.g., CXL+PRK or LASIK Xtra). FE models can simulate the increase in modulus after CXL and predict how much additional stabilization it provides. This guidance helps determine whether CXL is necessary and at what fluence.

Conclusion

Modeling the mechanical response of the eye to refractive surgery has matured from simple analytical formulas to sophisticated, patient-specific computational frameworks. Finite element analysis remains the gold standard, but machine learning surrogates and inverse methods are paving the way for clinical adoption. By accounting for individual tissue properties, IOP, and healing, these models enable surgeon to select safer ablation profiles, avoid ectasia, and optimize visual outcomes. Continued validation, integration into surgical systems, and coupling with optical simulations will further refine the art and science of refractive correction.

For further reading: