Purpose: To compare conventional two-dimensional multisection images with registered three-dimensional volume and subtraction images for detecting subtle changes in the brains of infants and children.
Methods: Twenty-six patients (24 with hemorrhagic/ischemic lesions) and one each with perinatal infection and Sturge-Weber disease were examined on two or more occasions with conventional multisection T1- and T2-weighted sequences as well as with 3-D T1-weighted volume sequences. A registration program was used to match the volume images to subvoxel dimensions, and subtracted images (second volume set minus the first) were obtained. The multisection images were compared with the 3-D and subtracted images and graded for detection of changes in a variety of brain structures.
Results: In 16% to 33% of comparisons of different structures, the multisection images and the 3-D registered and subtracted images showed changes equally well. The 3-D registered and subtracted images were better than the multisection images in 67% to 84% of comparisons for detection of changes in the cerebral hemispheres, ventricles, brain stem, cerebellum, and in lesions. Statistically significant differences were found between the graded performance of the registered 3-D images and the conventional 2-D images in detecting cerebral infarction and hypoxic ischemic encephalopathy. In the late phase following neonatal cerebral infarction (1 to 11 months), the 3-D registered and subtracted images revealed growth of the brain at the margins of the lesions.
Conclusion: Subvoxel registration of serial MR images may be of value in detecting subtle changes in the brains of infants and children.