A Regularized Two-tensor Model Fit to Low Angular Resolution Diffusion Images Using Basis Directions

Purpose: To resolve and regularize orientation estimates for two crossing fibers from images acquired with conventional diffusion tensor imaging (DTI) sampling schemes.

Materials And Methods: Partial volume causes artifacts in DTI. Given that routine use of high angular resolution diffusion imaging (HARDI) is still tentative, a regularized two-tensor model to resolve fiber crossings from conventional DTI datasets is presented. To overcome the problems of fitting multiple tensors, a model that exploits the planar diffusion profile in regions with fiber crossings is utilized. A regularization scheme is applied to reduce noise artifacts, which can be significant due to the relatively low number of acquired images. A set of basis directions is used to convert the two tensor model to many models of lower dimensionality. Relaxation labeling is utilized to select from amongst these models those that preserve continuity of orientations across neighbors. Revised fractional anisotropy (FA) and mean diffusivity (MD) values are computed.

Results: Spatial regularization improves the orientation estimates of the two-tensor model in simulations and in human data and estimates agree well with a priori anatomical knowledge.

Conclusion: Orientational, anisotropy, and diffusivity information can be resolved in regions of two fiber crossings using full brain coverage scans acquired in less than six minutes.