The effect of template selection on diffusion tensor imaging voxel based analysis results

Introduction: In recent years, voxel based analysis (VBA) studies have demonstrated the potential of diffusion tensor imaging (DTI) to detect white matter (WM) damage in patients with various neurological or psychiatric disorders. In VBA, all DTI data sets are transformed to an atlas or template, whereafter the diffusion measures of control subjects and patients are evaluated in each voxel. In most VBA studies of DTI data sets, a standard template, such as the Montreal Neurologic Institute (MNI) atlas, is used. The advantage of the MNI atlas is that it contains coordinate, anatomic, and cytoarchitectonic labels and that the VBA results can be easily compared across studies using the MNI coordinates. However, since this atlas is not study-specific, it might fail to provide a good representation of the population that is studied, thereby potentially resulting in residual image misalignment after coregistration of the DTI data sets to this reference space. Furthermore, since the original MNI template only contains anatomical MR or FA information, many studies only use the T1, T2, or FA as input information for the coregistration algorithm. This can affect the reliability of the VBA results, since it has been demonstrated that the use of tensor information during coregistration significantly improves the DT image alignment [1,2]. Recently a population based, study specific DTI atlas was introduced, whereby the magnitudes of the deformation fields that are needed to warp the different images to the atlas are minimized, potentially leading to a decreased image misalignment in VBA [3]. The goal of this work was to examine the effect of the atlas selection on the reported VBA results. To this end, the VBA sensitivity and specificity to detect WM damage is investigated using simulated DTI data sets [4]. The effect of two atlases on the VBA results is examined: a DTI atlas in MNI space and a population based, study specific DTI atlas, both including all DT information in the atlas space.