Modelling brain development to detect white matter injury in term and preterm born neonates.

摘要: Premature birth occurs during a period of rapid brain growth. In this context, interpreting clinical neuroimaging can be complicated by the typical changes in contrast, size and gyrification occurring background to any pathology. To model describe evolving shape we used Bayesian regression technique, Gaussian process regression, adapted multiple correlated outputs. Using MRI, simultaneously estimated tissue intensity on T1- T2-weighted scans as well local large cohort 408 neonates scanned cross-sectionally across perinatal period. The resulting provided continuous estimate intensity, appropriate age at scan, degree prematurity sex. Next, investigated utility detect focal white matter injury. individual neonates, calculated deviations neonate's observed MRI from that predicted punctate lesions with very good accuracy (area under curve > 0.95). investigate longitudinal consistency model, 46 who were second occasion. These infants' voxelwise could identify them other images 83% (T2-weighted) 76% (T1-weighted) cases, indicating an anatomical fingerprint. Our approach provides accurate estimates non-linear clear potential for radiological use.