Modified geodesic ray-tracing for diffusion tensor imaging

In this paper we develop a modified ray-tracing algorithm for geodesic tractography in the context of brain Diffusion Tensor Imaging (DTI). Our technique is based on computing multi-valued geodesics connecting two given points and tracking the evolution of adjacent geodesics. In order to do so we introduce a new Riemannian metric given by the adjugate sharpened diffusion tensor, combined with a constraint on the tracts outcome based on the geodesic deviation. We present tractography results, and compare our method with the existing ray-tracing approach and deterministic streamlining. Our preliminary results show an improved performance of modified ray-tracing regarding false positive fibers. We also show experiments on subcortical short association U-fibers, whose reconstruction is well-known to be hard in a DTI setting