Optimal diffusion tensor indices for imaging the human spinal cord

Diffusion tensor imaging (DTI) has been successfully used to image the human brain and spinal cord, although there is still controversy as to which tensor-derived diffusion indices produce the greatest contrast and provide the best anatomical representation of gray and white matter within the spinal cord. The aim of this study was to determine the best diffusion indices for use in the spinal cord using the detectability index, ROC analysis, and opinion data in the form of a survey. DTI of the entire spinal cord (C1-L1) was performed on five neurologically intact human subjects at 1.5-T. Eigenvalues, mean diffusivity (MD), fractional anisotropy (FA), volume ratio (VR), relative anisotropy (RA) and measured anisotropy (deviation of eigenvalues with respect to mean diffusivity) were calculated from the diffusion tensor. ROIs for white matter (WM) and gray matter (GM) were extracted using FA colormaps. The detectability index indicated FA and VR provided significant contrast between GM and WM. Since the FA was used for classification, results appeared to be biased toward FA and indices highly correlated with FA. ROC analysis illustrated similar results, but area under the ROC curve did not show statistical significance between indices. The survey indicated that the deviation of the primary eigenvalue with respect to mean diffusivity (MA1) was significantly better than all other indices at representing underlying spinal cord morphology. This is consistent with previous results showing lack of detail in ventral gray matter regions using the FA. Results indicate FA and MA1 provide the highest contrast and most accurate representation of underlying morphology, respectively.