Altered white brain matter structure in neonatal Ultrasound (US) images has prognostic implications for certain disorders. Commonly, physicians classify pathological white brain matter on a discrete categorical scale based on relevant qualitative characteristics. For certain pathologies, where subtle changes in structure have to be detected, this classification is too stringent. This is the case when characterizing affected white matter in the gliotic variant of Periventricular Leukomalacia (PVL), a brain disorder of very low birth weight preterm infants. The main objective of this study is to investigate quantitatively how texture information extracted from white matter regions in B-mode US images can guide physicians to a more accurate detection. A data set of 140 B-mode US images (70 non-pathological and 70 pathological) was investigated. Pathology was defined either by evolution to cystic PVL or by definite abnormality on acute MRI (ground truth). First, 7 different texture feature sets were extracted: First-Order statistics, Grey Level Co-occurrence matrix features, Run Length matrix features, Sum and Difference histogram features, Statistical features, Texture Energy Measure features and Gabor Filter features. Then, 3 classifiers were compared on these feature sets: a Bayesian Maximum A Posteriori (MAP) probability, a k Nearest Neighbor (kNN), and Fishers Linear Discriminant (FLD) classifier. Finally, a combination of the classifiers as well as texture feature combinations based on a confidence measure, were incorporated into a multi-feature, multi-classifier algorithm. Using our method, we succeeded in identifying the pathological group with an accuracy of 92.5% and sensitivity and specificity scores that exceed those of existing non-texture based methods. Consequently, this method can improve both the prognostic finesse and the guidance of early postnatal management.