Purchase this article with an account.
Anatoly Fabrikant; Objective Estimation for Uncertainty of Restoring Corneal Topography Surface. Invest. Ophthalmol. Vis. Sci. 2013;54(15):525.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Corneal topography (CT) field can be restored from measurements data by decomposing available data into Zernike polynomials and mapping the CT field in the desired area. The accuracy of such restoration depends on the measurement noise and number of available data. Here we present an algorithm for CT data assimilation, which yields both CT map and an objective estimation of the restoration accuracy.
The Kalman-Bucy technique is used to combine measured CT heights with a priori mean and covariance of Zernike coefficients, estimated for the general population (CT data for 308 virgin eyes). This algorithm yields a statistically optimal estimate of Zernike coefficients for the measured field and also their covariance matrix, which provides a measure of the measurement uncertainty. The efficiency of the proposed method is demonstrated using archived corneal topography data from previous clinical studies.
For any CT measurement the proposed technique yields an estimate of Zernike amplitudes and their covariance matrix, which result in a reconstructed map of CT heights with no gaps within the area (Fig.1A). The covariance matrix gives the uncertainty (std) of CT height restoration (Fig.1B). The uncertainty is higher at the area edges, because the restoration is based mainly on the data from internal area. Restored field in the measurement gaps has the highest uncertainty, close to the a priori variance of the general population.
The proposed algorithm assimilates measurement data together with a priori information, derived from statistics of general population, which protects the results from measurement outliers. It restores the CT field in the entire area and provides an objective estimate of measurement uncertainty, based on the measurement noise level and the number of available data. The uncertainty map displays the areas where the map is less reliable and to what extent.
This PDF is available to Subscribers Only