Anisotropy of the texture in the ultra-sonogram of human livers
Nivedita Neogi, Arunabha Adhikari, Madhusudan Roy · 2016
In the present study, discrimination between Ultra-sonograms USGs of human livers of fatty and normal kind, based on the difference of their image textures, is attempted. Different texture features have been suggested in the literature for this purpose. Most popular of them are GLCM based features. Other features have also been proposed for classification. However, a visual scrutiny of the ultra-sonograms indicates that a dominant characteristic of the textures in liver is the presence of small narrow line segments perpendicular to the motion of the beam line. For the most of the images these lines appear to run in nearly horizontal direction with increasingly higher radius of curvature in subsequent lower portions of the images. The texture of the image is therefore visually anisotropic. In this paper, we have first attempted to establish the anisotropy of the texture by statistical computations and proposed some features which explicitly take this directional asymmetry into account. We have extracted these new features from a data set comprising of human liver USG of normal and fatty class. Effectiveness of these new features is estimated by the p value of Welch's test.