Conclusions and Future Work

In this paper we developed a new technique for tracing anatomical fibers from 3D DT-MRI tensor fields, recovering identifiable anatomical structures that correspond to many of the white matter brain-fiber pathways. We found that simple component-wise interpolation of the tensors, forming a crude continuous approximate tensor field, worked well for extracting brain fiber directions when combined with a moving least squares filtering approach. Our plan is to extend the work to more advanced nonlinear filtering methods to better handle bifurcation points like crossings and ``T'' junctions. In addition, we plan to examine the benefits of detecting the local stiffness properties of the tensor field related to convergence or divergence of the fibers and automatically switch from forward to inverse integration schemes.