Deformation Based Morphometry with Implicit Reference-Based Registration: Validation and Detection of Structural Changes in a Primate Model of Early-Life Stress

IRG , respectively. The procedure was repeated with a set of different desired volume changes. The average Jacobian in the ROI was computed to provide the volume change obtained by DBM and simulation. The ratio of volume change (RVC) between DBM vs. simulation was also calculated. An RVC close to 1 indicates better performance. Another validation was to assess the sensitivity of the DBM method for detecting cross-group volume changes. Two groups of monkey brains were utilized. The left hippocampus (LHC) of each brain was manually segmented(4) and no significant volume difference between the two groups was found. For each image in the second group, a deformation field with a 10% atrophy (10% decrease of the original volume and Jac(h)=0.9) of the LHC was simulated using the same simulation tool, and was used to generate a new image with a smaller LHC. DBM was applied to the images of the first group and the second group before and after LHC shrinkage. T-tests were performed on the Jacobian maps to test for significant volume differences between the two groups before and after LHC shrinkage. Monkey brain structural change by DBM. 28 Rhesus monkeys assigned at birth either into mother-reared (MR) or peer-reared (PR) group were used, matched for gender (7 males and 8 females vs. 6 males and 7 females), age (27.4±0.9 vs. 26.4±0.3 months) and weight (3.6±0.13 vs. 3.5±0.1 kgs). MRI images were acquired on a 3T Siemens scanner and 224 slices were prescribed to cover the whole brain with a TR of 2.5 s, a TE of3.49 ms and a spatial resolution of 0.3×0.3×0.6 mm 3 . Data averaged from 4 runs were acquired using Nova DR dual surface coils. Affine registration was utilized to align each subject to a monkey brain template(7) to ensure that the overall brain volume of each subject was similar and the images were globally aligned. DBM was applied to generate Jacobian maps and voxel-wise t-tests performed. A threshold of corrected p<0.0 was used to generate the group difference map. 0.5 0.708 0.835 0.755 0.564 0.837 0.7 0.778 0.864 0.796 0.613 0.916 0.8 0.836 0.895 0.8418 0.638 0.964 0.9 0.909 0.940 0.908 0.665 1.02 0.95 0.954 0.969 0.953 0.673 1.02