FEATURE EVALUATION OF TEXTURE TEST OBJECTS FOR MAGNETIC RESONANCE IMAGING

Texture analysis of test object (phantom) images for standardization of in vivo magnetic resonance imaging is considered in this paper. The test objects are made of reticulated foam embedded in agarose gel. Different porosity foam materials are used to manufacture the phantoms. Both optical and MR images are analyzed, split into classes differing by the foam pore size. To characterize the image texture, a number of its firstand second-order statistical features is computed. The usefulness of the features to object class discrimination is evaluated using the ratio F of between-classes variance to within-classes variance, and multidimensional analysis of variance. The effect of noise and MR slice thickness on F is investigated.