Texture analysis within contrast enhanced abdominal CT images

The aim of this paper is to develop an automated texture based imaging system for the abdominal tissue classification within CT images. The development of such a system implies two major steps: the first step consists in finding the best features to describe the organs and the second one suppose a correctly classification of the desired tissue. In order to find relevant features, we realize a texture analysis on abdominal CT tissues by making use of a large variety of the most used texture features, including first and second order statistics, multiresolution analysis (Gabor filters) and Laws energy measures. Features of these types are used to classify three sets of different abdominal organs: liver, kidney and spleen. We also analyze the importance of the image contrast enhancement by making experiments on three different data sets: CT images in native phase, arterial phase and venous phase. Several classifiers were employed to build the classification system. Finally, we increase the classification system for this particular requirements by taking into account the CT capabilities of performing accurate geometrical localization of the anatomical structures.