Abstract
Purpose :
The elevation of the corneal surface can be obtained with anterior segment Optical Coherence Tomography (OCT). However, the lateral resolution obtained by a dense raster scan can be limited by the fact that the measurement is not instantaneous. We hypothesize that the simulation of fixational eye movements during the acquisition can predict a fraction of the variance found in the topographical measurements performed with these systems.
Methods :
Simulations were performed for two custom-built Fourier domain anterior segment OCT systems. In the first system (OCT1), a repetition rate of 25 KHz and 300x50 A-scans were used to capture volumetric images of the cornea in a 10x10 mm region. In the second system (OCT2), a repetition rate of 200 KHz and 300x150 A-scans/s with a waiting time between lines equivalent to 250 A-scans were used to image a 15x15 mm region. The measurement time was 0.6 and 0.41 s respectively. Lateral eye movements were simulated with a random-walk model (Engbert et al. 2011) and applied to a spherical cornea, r=7.9 mm. On each scanner position, eye movement was applied and the sag calculated. The central 6 mm of the data were fitted by a sphere (radius R) and the residuals were fitted by Zernike polynomials (6th order coefficients, zC). Experimental tests were performed on normal subjects (10 for OCT1, 4 for OCT2) and the corneal elevation fitted with the same algorithms. The standard deviation (std) of the fitting parameters in simulation was compared to the experimental.
Results :
The random-walk model shifted the cornea a maximum of 200 μm. The repeated simulations with different eye movements predicted a std of 0.15 and 0.12 mm in R and a std of 3.8 and 2.9 µm for the third and 2.5 and 1.5 µm for the fourth order zC in OCT1 and OCT2 respectively. The repeated measurements in subjects showed an average std across subjects of 0.18 and 0.19 mm for R and 3.7 and 3.6 for the third and 2.4 and 1.8 µm for the fourth order zC in OCT1 and OCT2 respectively.
Conclusions :
Our simulations of the raster scan acquisition of corneal topographies including lateral eye movements predicts a std in R that is 70% of the std found experimentally. Other sources of variance such as rotations, axial movements or instrument and image processing precision were not considered. However, these simulations can serve to understand the sources of variability in topography measurements and to evaluate correction algorithms.
This is a 2020 ARVO Annual Meeting abstract.