Texture Analysis in 3D for the Detection of Liver Cancer in X-ray CT Scans
Manos Papadakis, University of Houston
Location: Stevenson 1307
We propose a method for the 3D-rigid motion invariant texture discrimination for discrete 3D-textures that are spatially homogeneous. We model these textures as stationary Gaussian random fields. We formally develop the concept of 3D-texture rotations in the 3D-digital domain. We use this novel concept to define a `distance' between 3D-textures that remains invariant under all 3D-rigid motions of the texture. This concept of `distance' can be used for a monoscale or a multiscale setting to test the 3D-rigid motion invariant statistical similarity of stochastic 3D-textures. To extract this novel texture `distance' we use the Isotropic Mutliresolution Analysis. We also show how to construct wavelets associated with this structure by means of extension principles and we discuss some very recent results by Atreas, Melas and Stavropoulos on the geometric structure underlying the various extension principles. The 3D-texture `distance' is used to define a set of 3D-rigid motion invariant texture features. We experimentally establish that when they are combined with mean attenuation intensity differences the new augmented features are capable of discriminating normal from abnormal liver tissue in arterial phase contrast enhanced X-ray CT-scans with high sensitivity and specificity. To extract these features CT-scans are processed in their native dimensionality. We experimentally observe that the 3D-rotational invariance of the proposed features improves the clustering of the feature vectors extracted from normal liver tissue samples. This work is joint with R. Azencott, S. Jain, S. Upadhyay, I.A. Kakadiaris and G. Gladish, MD.