Optimized Voxel-Based Group Analysis of Diffusion Tensor Imaging by means of Barycentric Linear Anisotropy

Introduction: Diffusion tensor imaging (DTI) is a non-invasive magnetic resonance method providing a measure of organization of white matter. It has gained increasing popularity to study disorders involving a potential neurodevelopmental origin and disruption of myelination. Employing DTI, altered integrity of the corpus callosum and frontostriatal tracts were reported in schizophrenia. The common quantification of anisotropy for voxel-based group analysis is the fractional anisotropy. This coefficient may be overestimated with low signal-to-noise ratios and may be inappropriate in situations of complex fiber crossing, compartimentalization, exchange of water and in subcortical white matter. To partially circumvent these limitations, the present study aimed at examining the potential of the barycentric representation composed of three complementary anisotropy coefficients describing the following diffusion shapes: linear (e.g. axonal fibers), planar (e.g. crossing structures) and spherical (e.g. purely Gaussian water diffusion). Method: 30 schizophrenic patients (mean age 27.5 years) and 18 healthy volunteers (mean age 28.7 years), matched for age, gender, and education, were measured by DTI on a 1.5T Siemens Magnetom (Vision): Double spin-echo EPI (Hunsche 2001), b-value of 900s/mm², six independent directions, TE/TR=100 ms/ 4000 ms, 4 repetitions, 0.9375×0.9375mm² in-plane resolution,128×128 zero-filled, 240mm FOV, 3mm slice thickness, two acquisitions of 116s, 38 slices. Diffusion tensor and linear anisotropy processing routines were programmed under MATLAB. Results and Discussion: Linear anisotropy was employed to filter possible crossing structures (planar anisotropy) and compared to fractional anisotropy. At lower signal-to-noise ratios linear anisotropy maps provided better defined boundaries at the edge of white and grey matter. The apparent lesser vulnerability of linear anisotropy to signal-to-noise and its higher specificity to linearly oriented diffusion may provide a more robust basis for voxel-based group analysis of white matter changes near the cortex. This may be particularly important to detect alterations of medium range cortico-cortical connectivity fibers.

Support: BMBF IZKF FKZ01ZZ0105, TMWFK B30701–015/-016