Tracking of axonal bundles in diffusion MRI brain images

The aim of this thesis is to design tracking algorithm which will be able to track white matter bundles in diffusion MRI data, this problem is called tractography. Tractography is feasible because specific profile of diffusion appears in white matter. The introduction to the topic includes summary of methods for estimation of diffusion profile and basic tracking algorithms. In this work diffusion tensor model (DTI) was used for estimation of diffusion profile. Based on the DTI, vector field characterizing direction of diffusion for every voxel was created. Combining vector field with seedpoint, we achieved task solvable by Euler or Runge-Kutta method. Termination criteria were established for maximum curvature of trajectory and minimum value of fractional anisotropy (FA). Algorithm was tested on mathematical and tractographical phantom before it was used on real biological data. The results of tracking on phantoms proved the funcionality of the algorithm. Expected error appeared in areas of crossing fibers, it is related to DTI model limitations. To solve problematic fibers characterized by seedpoint near the border of the fiber, FA-weighted trilinear interpolation was designed. Implementation of this algorithm, however, did not cause better results. The results of tracking on the real data were controversial. Tracking was performed on 5 healthy subjects and 4 anatomicaly specific tracts. The results were compared with tractographic atlas