Objective Diffuse white matter injury (WMI) among preterm infants is associated with deep grey matter volume loss. Computational techniques that identify subtle structural alteration involve thousands of simultaneous voxelwise tests, resulting in unacceptably high type 1 error. One correction method limits the number of false positives throughout the entire volume, familywise error rate (FWER). A new more powerful approach is to control the proportion of false positives among the total number of positives (false discovery rate; FDR). Hypothesis: additional morphological changes attributable to diffuse WMI are detected by controlling FDR in group analyses.
Methods 89 preterm infants at term without major parenchymal lesions and 20 term controls underwent volumetric T1 and diffusion-weighted magnetic resonance imaging at 1.5 Tesla. Structural data were aligned to a template using non-rigid registration and volume change maps were calculated. These were analysed by t-test with (1) FWER and (2) FDR control (SPM5). Diffuse WMI was defined quantitatively and infants with (n = 66) and without (n = 14) WMI were compared with the control group.
Results Diffuse WMI is associated with tissue loss within the corona radiata and periventricular white matter detected using FDR but not FWER control (t = 4.42, p = 0.001) (fig A and B). Infants with diffuse WMI had tissue loss within the globus pallidus and dorsomedial complex of the thalamus (p = 0.001) (fig B), whereas those with normal white matter did not.