Home Research Project Details C2 - Numerical methods for MRI-based fiber tractography of the brain
Personal tools

C2 - Numerical methods for MRI-based fiber tractography of the brain

Jens Frahm and Thorsten Hohage

Diffusion-weighted MRI of the brain allows for the virtual reconstruction of the white matter fiber architecture and thus yields structural connectivities that directly subserve the inter-areal communication between brain systems. So far, our work focused on the development of an MRI technique (STEAM) that maps diffusion parameters without geometric distortions [Mori et al. 1999]. The use of a diffusion tensor model and a simple algorithm for fibre tracking [Nolte et al. 2000] then led to successful neurobiological applications that unravelled major fibre pathways in the brains of humans [Hofer et al. 2006, Hofer et al. 2008], rhesus monkeys [Hofer et al. 2008, Hofer, Frahm 2008], and (genetically modified) mice [Boretius et al. 2009]. However, current methods still suffer from limitations and, for example, are unable to deal with “crossing” or “kissing” fibre tracts in individual image voxels. To overcome these limitations we plan to address the following problems:

(a) Data acquisition: Because the spatial resolution and achievable SNR remain key obstacles, we will further improve the MRI acquisition. Promising trials are due to a segmented MRI technique and the estimation of motion-induced phase maps at high resolution [Uecker et al. 2009] with the use of a previously developed method defining the reconstruction as a nonlinear inverse problem [Uecker et al. 2008]. This work will be extended to non- Cartesian trajectories, which offer even more efficient data sampling strategies.

(b) Diffusion modelling: The simple tensor model does not meet the complex behaviour of water mobility in brain tissue. For this reason, more refined models with higher angular resolution have to be employed (e.g., [Tuch et al. 2002]) that in turn require new numerical analysis methods (e.g., [Franken et al. 2009]). In addition, we plan to develop model-based reconstruction techniques from nonlinear inverse problems to directly estimate diffusion properties from the MRI data without intermediate image reconstruction.

(c) Fibre tracking: In order to exploit the information of a more realistic diffusion model, new algorithms for fibre tracking are required. The development of probabilistic approaches, which do not track single fibers but estimate the likelihood of connections between brain areas, represent one possibility to deal with the statistical uncertainty of the data. These algorithms are also expected to allow for a better regularization by the incorporation of a priori biological information.

(d) The combined optimization of a suitable MRI technique, a reasonable diffusion model, and an adequate algorithm for fibre tractography requires a tight connection and inter-relationship to the biological questions posed in humans, nonhuman primates, and rodents. Based on our previous experience in these areas (partly gained through work in BCCN I) we expect to benefit from and be helpful to almost all other subprojects in section C. In addition, the results may be of significant relevance for translational research from mouse to human and of direct clinical applicability.

Belongs to Group(s):
NMR in living systems,

Is part of  Section C 

Members working within this Project:
Frahm, Jens 
Hohage, Thorsten  

Selected Publication(s):