Abstract
Purpose :
This research aims to assess the accuracy of patient-specific finite element (FE)-based modelling for predicting corneal geometric responses to LASIK.
Methods :
Retrospectively, baseline corneal tomography (Pentacam) data from 11 eyes of 7 subjects (age: 43.0 ± 10.3 years, mean spherical equivalent refractive error treated: -2.61D, range: -1.00 to -4.50D) were used to generate eye models with patient-specific corneal geometries. Point cloud data from tomography was reconstructed using Zernike fits and the model was discretized with 20 node hexahedral elements with in-house meshing code. Then, custom FE code with a microstructurally motivated material model and prestressing algorithm was used to simulate the LASIK (flap and myopic photoablation) procedure. In the current study, a multi-scale approach was employed to capture the anisotropic and non-linear corneal response. Anterior tangential and axial curvatures as well as elevation maps were computed for the central circular area (3, 5, and 8mm diameter). FE-derived values were compared to preop (verification) and postop (validation) clinical tomography to evaluate the predictive performance of the computational models.
Results :
The differences in preop shape (FE derived - clinical values) were as follows: for tangential curvature, -0.01±0.35 D (3mm), -0.01±0.41 D (5mm) and -0.01±0.64 D (8mm); for axial curvature, 0.05±0.37 D (3mm), 0.03±0.31 D(5mm) and 0.02±0.25 D (8mm); for elevation, -0.2±0.4 µm (3mm), -0.3±0.4 µm(5mm) and -0.3±0.4 µm (8mm). Prediction errors for post-Lasik data were as follows: for tangential curvature, -0.36±1.13 D (3mm), -0.61±1.22 D (5mm) and -0.13±1.73 D (8mm); for axial curvature, 0.03±1.24 D (3mm), -0.20±1.09 D (5mm) and -0.27±0.93 D (8mm); for elevation, -0.2±2.0 µm (3mm), 0.3±2.2 µm (5mm) and 0.3±2.7 µm (8mm).
Conclusions :
FE models with patient-specific corneal geometries can predict LASIK outcomes with a mean error within the limits of clinical significance, potentially providing valuable insights for clinicians in treatment planning. Variance in curvature predictions may improve by incorporating patient-specific material properties.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.