The developmental significance of the frequently encountered white matter signal abnormality (WMSA) findings on MRI around term-equivalent age (TEA) in very preterm infants, remains in question. The use of conventional qualitative analysis methods is subjective, lacks sufficient reliability for producing accurate and reproducible WMSA diagnosis, and possibly contributes to suboptimal neurodevelopmental outcome prediction. The advantages of quantitative over qualitative diagnostic approaches have been widely acknowledged and demonstrated. The purpose of this study is to objectively and accurately quantify WMSA on TEA T2-weighted MRI in very preterm infants and to assess whether such quantifications predict 2-year language and cognitive developmental outcomes. To this end, we constructed a probabilistic brain atlas, exclusively for very preterm infants to embed tissue distributions (i.e. to encode shapes, locations and geometrical proportion of anatomical structures). Guided with this atlas, we then developed a fully automated method for WMSA detection and quantification using T2-weighted images. Computer simulations and experiments using in vivo very preterm data showed very high detection accuracy. WMSA volume, particularly in the centrum semiovale, on TEA MRI was a significant predictor of standardized language and cognitive scores at 2 years of age. Independent validation of our automated WMSA detection algorithm and school age follow-up are important next steps.