Abstract
Abstract: :
Purpose: To investigate the effects of noise, sampling schemes, and reconstruction on the estimation of the wave aberrations of the eye. Methods: A software suite was developed in MATLAB (Mathworks, Inc, Natick MA) that allowed the measurement process for wave aberrations to be simulated. Using this suite arbitrary sampling schemes could be tested, and the efficiency and accuracy of the reconstructed wavefronts could be directly compared to the original wavefronts. In addition, known amounts of noise could be introduced to test the resilience of the sampling schemes to noise. Data from both a Spatially Resolved Refractometer and a Hartmann Shack sensor were used to constrain models, and compare possible tradeoffs in sampling. Results: While there were consistent errors in identifying Zernike terms (a form of aliasing), these were quite small (< 0.06 µm) when centroiding noise was set at 0.03 mR per sample. In general the accuracy of locating the centroid was the major factor contributing to accurate determination, as long as there were more points measured than there were Zerrnike terms being reconstructed. Increasing the sampling density at the expense of noise while keeping the number of Zernike terms constant was not beneficial. That is, if the centroiding error increases inversely with the area of the pupil being sampled, then the benefit of more samples is eliminated by the increased noise in the centroiding process. For sparse sampling (as can be done in sequential techniques) non-regular sampling schemes (cubature, etc) were somewhat more efficient than grid sampling, but only when sampling noise was higher than is likely to be the case for most measurements in the eye. In conditions where there were localized defects in the wavefront (eg, due to surgery, tear film beneath the lid, or partial obscuration from the pupil margin) it was critical that all orders of the polynomial fit be included in the reconstruction, or else the local nature of the change was lost in the reconstruction. Conclusions: Sampling density is not a major problem in estimating wave aberrations unless very high order fits are being considered. Centroiding noise is critical, and there is a tradeoff between number of samples and noise. Different sampling schemes can cause systematic errors in identifying specific combinations of Zernike modes.
Keywords: optical properties • physiological optics • refractive surgery