Anisotropic texture modeling and applications to medical image analysis
Hermine Biermé, Frederic J. P. Richard, M. Rachidi, Claude-Laurent Benhamou · ESAIM Proceedings · 2009
In this paper, we consider a stochastic anisotropic model for medical image. We model textures by Anisotropic Fractional Brownian Fields (AFBF) which are Gaussian random fields obtained as anisotropic generalizations of the Fractional Brownian Field. The main difficulty with this modeling consists in the estimatation of the anisotropy. We recall here theoretical results obtained in [2,5] to construct consistent estimators for the texture anisotropy analysis. These results allow us to propose statistical tests for deciding if a texture modeled by an AFBF is clearly anisotropic. We apply them to Full-Field Digital Mammograms and show that about 60 percent of textures can be considered as anisotropic with a high level of confidence. We finish this paper with preliminary results on trabecular bone x-ray images. In [4], a fractal analysis based on isotropic Fractional Brownian Fields was proposed to characterize bone microarchitecture. Our first results suggest that anisotropic models are also relevant for the statistical analysis of this microarchitecture.