Statistical analysis of corneal OCT speckle: a non-parametric approach

In biomedical optics, it is often of interest to statistically model the amplitude of the speckle using some distributional approximations with their parameters acting as biomarkers. In this paper, a paradigm shift is being advocated in which non-parametric approaches are used. Specifically, a range of distances, evaluated in different domains, between an empirical non-parametric distribution of the normalized speckle amplitude sample and the benchmark Rayleigh distribution, is considered. Using OCT images from phantoms, two ex-vivo experiments with porcine corneas and an in-vivo experiment with human corneas, an evidence is provided that the non-parametric approach, despite its simplicity, could lead to equivalent or better results than the parametric approaches with distributional approximations. Concluding, in practice, the non-parametric approach should be considered as the first choice to speckle modeling before a particular distributional approximation is utilized.