Seiji Kumazawa PhD, Presenter: Nothing to Disclose

Takashi Yoshiura MD, Abstract Co-Author: Nothing to Disclose

Futoshi Mihara MD, Abstract Co-Author: Nothing to Disclose

Hiroshi Honda MD, Abstract Co-Author: Nothing to Disclose

Yoshiharu Higashida PhD, Abstract Co-Author: Nothing to Disclose

Fukai Toyofuku PhD, Abstract Co-Author: Nothing to Disclose

Conventional fiber tracking algorithms follow a path of white matter tract in a point-to-point manner based on the principle direction of anisotropic diffusion tensors (DTs). This approach fails to trace the path at the crossing or branching of the fiber tracts. Our goal was to develop a new white matter tractography algorithm in DT-MRI which permits fiber tracts crossing and branching.

To overcome the above problem in conventional algorithms, we introduced a 3D density field of the white matter tracts using the metaball technique. To express the distribution of tract density based on the geometric nature of the DT within a voxel, we defined a polarized metaball which had a 3D Gaussian distribution of positive or negative density. Using a self-organizing map, voxels in DT-MRI were classified into three clusters according to the geometric measures of DT as follows; 1) linear diffusion (LD) cluster, 2) planar diffusion (PD) cluster, and 3) spherical diffusion (SD) cluster. Positive and negative metaballs were generated on each voxel of LD cluster and SD cluster, respectively. No metaball was assigned to the PD cluster. By summating effects of all metaballs in the voxel space, a map of density field of tracts was generated. In this tract density field, our algorithm tracked voxels with a density larger than a certain threshold value. We tested the algorithm on both human brain DT-MRI and synthetic datasets. The human brain data consisted of 6 diffusion-weighted image volumes (b=800 s/mm2) and an unweighted image volume (b=0 s/mm2) with a 128x128 in-plane resolution. The synthetic datasets contained crossing and branching fibers which simulated the human brain DT datasets.

Our algorithm correctly followed tracts through crossing and branching on synthetic data. In the human brain datasets, major white matter tracts could be tracked beyond the fiber crossing.

We developed a new tractography algorithm which permits tracking through crossing and branching. Efficacy of our new method was demonstrated in synthetic data and human brain data.

Kumazawa, S, Yoshiura, T, Mihara, F, Honda, H, Higashida, Y, Toyofuku, F, White Matter Tractography with Metaball Technique in Diffusion Tensor MRI.  Radiological Society of North America 2004 Scientific Assembly and Annual Meeting, November 28 - December 3, 2004 ,Chicago IL. http://archive.rsna.org/2004/4406526.html