Spatial normalization and regional assessment of cord atrophy: Voxel-based analysis of cervical cord 3D T1-weighted images

BACKGROUND AND PURPOSE: VBM is widely applied to characterize regional differences in brain volume among groups of subjects. The aim of this study was to develop and validate a method for voxelwise statistical analysis of cord volume and to test, with this method, the correlation between cord tissue loss and aging. MATERIALS AND METHODS: 3D T1-weighted scans of the spinal cord were acquired from 90 healthy subjects spanning several decades of life. Using an AS method, we outlined the cord surface and created output images reformatted with image planes perpendicular to the estimated cord centerline. Unfolded cervical cord images were coregistered into a common standard space, and smoothed cord binary masks, produced by using the cord outlines estimated by the AS approach, were used as input images for spatial statistics. RESULTS: High spatial correlation between normalized images was observed. Averaging of the normalized scans allowed the creation of a cervical cord template and of a standardized region-of-interest atlas. VBM analysis showed some significant associations between a decreased probability of cord tissue and aging. Results were robust across different smoothing levels, but the use of an anisotropic Gaussian kernel gave the optimal trade-off between spatial resolution and the requirements of the Gaussian random field theory. CONCLUSIONS: VBM analysis of the cervical cord was feasible and holds great promise for accurate localization of regional cord atrophy in several neurologic conditions.